Apparatus and methods for including codes in audio signals and decoding

The present invention relates to apparatus and methods for inserting codes into audio signals and for decoding such codes.

Techniques have been proposed for many years to mix codes with audio signals so that (1) the codes can be reliably reproduced from the audio signals, while (2) the codes are inaudible when the audio signals are reproduced as sound. Compliance with these two conditions is essential for practical purposes. For example, broadcasters and broadcast producers, as well as recording companies of music for public performances, would not tolerate the addition of audible codes to their programs or recordings.

Such techniques for coding audio signals have already been proposed at different points in time and go at least as far as LI.S. 3,004,104 (Hembrooke, issued October 10, 1961). In said patent an encoding method is shown in which the energy of the audio signal has been selectively removed within a narrow frequency band in order to encode the signal. A problem with this technique is that the code becomes indistinct when noise or signal distortion re-energizes this narrow frequency band.

In another method according to U.S. Patent No. 3,845,391 (Crosby) it is proposed to cut a narrow frequency band out of the audio signals and insert a code there. Obviously, this technique has the same problems as above as in US Pat. 4,703,476 (Howard), which it can be seen to be filed by the same assignee as the Crosby patent. The Howard method, however, merely sought to improve the Crosby method without departing from its basic approach.

It has also been proposed to encode binary signals by distributing the binary codes at frequencies extending over the entire audio band. A problem with this method is that the code frequencies can become audible if there are no audio frequency components to obscure them. This method is therefore based on the assumption that the codes have a noise-like character and it can thus be assumed that their presence will not be perceived by the listeners. In many cases, however, this prerequisite is not correct, for example when, in the case of classical music, there are periods of time with a relatively low audio signal content, or during pauses in speech.

Another technique has been proposed in which dual tone multifrequency (DTMF) codes are inserted into an audio signal. The DTMF codes are allegedly detected based on their frequencies and durations. However, audio signal components can be mistakenly taken for one or both tones of each DTMF code, so that either the presence of a code is not recognized by the detector or, conversely, signal components are mistakenly mistaken for a DTMF code. In addition, it should be noted that each DTMF code has a tone which it has in common with another DTMF code. Accordingly, a signal component corresponding to a tone of another DTMF code may combine with the tone of a DTMF code coexisting in the signal, resulting in erroneous detection.

Furthermore, US Pat. No. 4,771,455 (Hareyama et al.) Discloses a transmitter and a receiver for the common transmission of a main signal and a scrambled signal by time multiplication. The receiver separates the two signals based on their time-division multiplex relationship. While the scrambled signal will not be transmitted if the level of the main signal is below a level V._S.h falls, the receiver simply writes the received scrambled signal into the memory at successive addresses and outputs data from this data pool. Assigning an amplitude to the scrambled signal on the basis of the level of the main signal is not apparent from US Pat. No. 4,771,455. Rather, the transmitter does not insert a scrambled signal into the main signal if the level of the main signal falls below a predetermined value V._SH falls.

It is known from US Pat. No. 5,213,337 to encode information in code tones in the audio part of a TV program and to mask the code tones by sound effects, with either a technician deciding where the codes for code tones are inserted or these

AT 410 047 B can be automatically inserted into the soundtrack with the sound effects.

In U.S. Patent 4,876,617, a binary code is inserted into two frequency notches of an audio signal. Best et al. try to cover the inserted codes by adjusting their amplitudes based on the envelope of part of the audio signal in a range from 1 kHz to 6 kHz. The codes are detected simply by bandpass filtering and obviously by threshold value detection.

Finally, WO 93/07689 (Arbitron) shows an audio signal encoder which encodes predetermined narrow frequency bands of an audio signal either by reducing the energy in the frequency bands to zero or by increasing it to a predetermined value.

The object of the present invention is therefore to provide devices and methods for coding and decoding which overcome the aforementioned and further disadvantages of the known techniques explained above.

A further object of the invention is to provide devices and methods for inserting codes into audio signals, so that the codes are inaudible as a tone to the human ear, but can be reliably detected by a decoding device. Finally, the object of the present invention is furthermore to provide coding devices and methods in order to reliably restore codes which are present in audio signals.

According to a first embodiment of the present invention, the device and method for inserting a code with at least one coding frequency component into an audio signal with a number of audio signal frequency components comprise the following steps or Means for carrying out the same: evaluating the suitability of a first set of the plurality of audio signal frequency components for masking the at least one code frequency component from the human ear in order to generate a first masking evaluation; Evaluating the suitability of a second set of the plurality of audio signal frequency components that differ from the first set to mask the at least one code frequency component from the human ear to generate a second masking score; Assigning an amplitude to the at least one code frequency component based on a selected first or second masking score; and inserting the at least one code frequency component into the audio signal.

In another embodiment of the invention, an apparatus for inserting a code with at least one code frequency component into an audio signal with a number of audio signal frequency components includes: a digital computer having an input for receiving the audio signal, the digital computer being programmed to evaluate the respective suitability of the first and second sets of the number of audio signal frequency components with respect to which at least one code frequency component is obscured from the human ear, to generate a first and a second masking assessment, wherein the second set of the number of audio signal frequency components is different from the first set, and wherein the digital computer is further programmed to assign an amplitude to the at least one code frequency component based on a selected first or second masking score; and means for inserting the at least one code frequency component into the audio signal.

In a further embodiment of the invention, devices and methods for inserting a code with a number of code frequency components in an audio signal with a number of audio signal frequency components, the number of code frequency components being a first code frequency component with a first frequency and a second code frequency component with a second frequency different from the first frequency, the following steps or Means for carrying it out on: evaluating the suitability of at least one of the plurality of audio signal frequency components for masking a code frequency component having the first frequency from the human ear in order to generate a first masking evaluation; Evaluating the suitability of at least one of the plurality of audio signal frequency components for masking a code frequency component having the second frequency from the human ear in order to carry out a second masking evaluation, assigning an amplitude to the first code frequency component on the basis of the first masking evaluation and Assign an amplitude to the

AT 410 047 B second code frequency component based on the second masking score; and inserting the number of code frequency components into the audio signal.

According to a further embodiment of the invention, devices and methods for inserting a code with a number of code frequency components in an audio signal with a number of audio signal frequency components, the number of code frequency components being a first code frequency component with a first frequency and a second code frequency component with a second frequency different from the first frequency, the following steps or Means for their execution on: a digital computer having an input for receiving an audio signal, the digital computer being programmed to assess the suitability of at least one of the number of audio signal frequency components for obscuring a code frequency component having a first frequency from the human ear to a carry out the first concealment assessment, and to evaluate the suitability of at least one of the plurality of audio signal frequency components for masking a code frequency component having a second frequency from the human ear to generate a second masking score; wherein the digital computer is further programmed to assign a corresponding amplitude to the first code frequency component based on the first masking score and to assign a corresponding amplitude to the second code frequency component based on the second masking score; and means for inserting the number of code frequency components into the audio signal.

In a further embodiment of the invention, methods and devices for inserting a code with at least one code frequency component into an audio signal with a number of audio signal frequency components include the following steps or Means for their implementation: Assessment of the suitability of at least one of the number of audio signal frequency components within a first audio signal interval on a time scale of the audio signal when reproduced as a tone during a corresponding first time interval for concealing the at least one code frequency component from the human ear, when reproduced as a tone during a second time interval corresponding to a second audio signal interval, offset from the first audio signal interval to generate a first masking score; Assigning an amplitude to the at least one code frequency component based on the first masking score; and inserting the at least one code frequency component into a portion of the audio signal within the second audio signal interval.

In a further embodiment of the invention, a device for inserting a code with at least one code frequency component into an audio signal with a number of audio signal frequency components includes: a digital computer with an input for recording the audio signal, wherein the digital computer is programmed to evaluate the suitability of at least one of the number of audio signal frequency components within a first audio signal interval on a time scale of the audio signal when reproduced as a sound during a first time interval for obscuring the at least one code frequency component from the human ear Playback as sound during a second time interval corresponding to a second audio signal interval, offset from the first audio signal interval to generate a first masking score; wherein the digital computer is further programmed to assign an amplitude to the at least one code frequency component based on the first masking score; and means for inserting the at least one code frequency component into a portion of the audio signal within the second audio signal interval.

In a further embodiment of the invention, devices and methods for inserting a code with at least one code frequency component into an audio signal with a number of audio signal frequency components include the following steps or the means for carrying it out: generating a first audio signal which essentially represents a first, individual one of the number of audio signal frequency components; Evaluating the suitability of the first, individual one of the plurality of audio signal frequency components for concealing the at least one code frequency component from the human ear on the basis of the first audio signal in order to generate a first concealment evaluation; Assigning an amplitude to the at least one code frequency component based on the first masking score;

AT 410 047 B and inserting the at least one code frequency component into the audio signal.

In a further embodiment of the invention, a device for inserting a code with at least one code frequency component into an audio signal with a number of audio signal frequency components includes: a digital computer having an input for receiving the audio signal, the digital computer being programmed to generate a first audio signal that is substantially a first, individual one of the plurality of audio signal frequency components, and for evaluating a suitability of the first, individual one from the number of audio signal frequency components for concealing the at least one code frequency component from the human ear, on the basis of the first audio signal, to generate a first occlusion score; wherein the digital computer is further programmed to assign an amplitude to the at least one code frequency component based on the first masking score; and means for inserting the at least one code frequency component into the audio signal.

In another embodiment of the invention, apparatus and methods for detecting a code in an encoded audio signal, the encoded audio signal having a number of audio frequency signal components and at least one code frequency component having an amplitude and an audio frequency selected to obscure the code frequency component from the human hearing through at least one of the number of audio frequency signal components, the following steps or means for carrying them out: establishing an expected code amplitude of the at least one code frequency component on the basis of the encoded audio signal; and detecting the code frequency component in the encoded audio signal based on its expected code amplitude.

In a further embodiment of the invention there is a programmed digital computer for detecting a code in an encoded audio signal, the encoded audio signal having a number of audio frequency signal components and at least one code frequency component having an amplitude and an audio frequency selected to mask the code frequency component the human ear through at least one of the number of audio frequency signal components, the digital computer comprising: an input for receiving the encoded audio signal; a processor programmed to establish an expected code amplitude of the at least one code frequency component based on the encoded audio signal, to detect the code frequency component in the encoded audio signal based on the expected code amplitude, and to generate a code detection output signal on the Generate the basis of the detected code frequency component; and an output connected to the processor for providing the code detection output signal.

In a further embodiment of the invention, methods and devices for detecting a code in an encoded audio signal include wherein the encoded audio signal has a number of frequency components including a number of audio frequency signal components and at least one code frequency component having a predetermined audio frequency and a predetermined amplitude for distinguishing the at least one code frequency component from the number of audio frequency signal components, the following steps or Means for carrying it out: determining an amplitude of a frequency component of the encoded audio signal within a first range of audio frequencies including the predetermined audio frequency of the at least one code frequency component; Establishing a noise amplitude for the first range of audio frequencies; and detecting the presence of the at least one code frequency component in the first range of audio frequencies based on its established noise amplitude and the determined amplitude of the frequency component therein.

In a further embodiment of the invention there is a digital computer for detecting a code in an encoded audio signal, wherein the encoded audio signal has a number of frequency components including a number of audio frequency signal components and at least one code frequency component having a predetermined audio frequency and a predetermined amplitude for distinguishing the at least one code frequency component from the number of audio frequency signal components, comprising: an input for receiving the encoded audio signal; a processor connected to the input to the

AT 410 047 B

Receiving the encoded audio signal programmed to determine an amplitude of a frequency component of the encoded audio signal within a first range of audio frequencies including the predetermined audio frequency of the at least one code frequency component; wherein the processor is further programmed to establish a noise amplitude for the first range of audio frequencies and to detect the presence of the at least one code frequency component in the first range of audio frequencies based on the established noise amplitude thereof and the established amplitude of the frequency component therein; wherein the processor is operable to generate a code output signal based on the detected presence of the at least one code frequency component; and having an output terminal connected to the processor at which the code signal is provided.

In a further embodiment of the invention, devices and methods for coding an audio signal include the following steps or Means for their implementation: Generating a code having a number of sets of code frequency components, each of the sets of code frequency components representing a different code symbol and including a number of different code frequency components, the code frequency components of the sets of code frequency components forming groups that are spaced apart within the frequency range, wherein each of the component groups has a predetermined frequency range and consists of a frequency component of each of the set of code frequency components falling within its respective predetermined frequency range, component groups which are adjacent within the frequency range being separated from one another by a respective frequency amount, the predetermined frequency range of each component group is smaller than the frequency amounts, which separate the respective component group from its neighboring component groups; and combining the code with the audio signal.

In a further embodiment of the invention there is a digital computer for encoding an audio signal, comprising: an input for receiving the audio signal, a processor which is programmed to generate a code with a number of sets of code frequency components, each set of the code frequency components representing a respectively different code symbol and a number of respectively different code frequency components includes, wherein the code frequency components of the sets of code frequency components form component groups, which are spaced apart within the frequency range, each of the component groups having a predetermined frequency range and consisting of a frequency component of each of the set of the code frequency components falling within their respective predetermined frequency range, component groups which are adjacent within the frequency range by a respective one Frequency amount are separated from each other, wherein the predetermined frequency range of each component group is smaller than the frequency amounts which separate the respective component group from its neighboring component groups; and means for combining the code with the audio signal.

The invention is explained in more detail below with the aid of preferred exemplary embodiments and with reference to a drawing. It shows:

Fig. 1 is a block diagram for the function of a coding device according to a

Aspect of the present invention;

Fig. 2 is a block diagram showing the function of a digital coding apparatus according to an embodiment of the present invention;

Figure 3 is a block diagram of a coding system for use in coding audio signals that are in analog form;

4 shows spectral representations to explain the frequency composition of different data symbols, as they are coded by the embodiment according to FIG. 3;

FIGS. 5 and 6 are functional block diagrams for explaining the operation of the embodiment according to FIG. 3;

7A to 7C each show a flow chart for explaining a software program which is used in the embodiment according to FIG. 3;

AT 410 047 B

7D and 7E each show a flow chart for explaining an alternative software program which is used in the embodiment according to FIG. 3;

Fig. 7F is a graph showing a linear approximation of a single tone masking relationship;

Fig. 8 is a block diagram of a coding apparatus employing analog circuitry;

FIG. 9 is a block diagram of a circuit for determining a weighting factor in the embodiment of FIG. 8;

Figure 10 is a functional block diagram of a decoding apparatus incorporating certain features of the present invention;

11 is a block diagram of a decoding apparatus according to an embodiment of a present invention using digital signal processing;

FIGS. 12A and 12B each show a flow chart for describing the operation of the decoding device according to FIG. 11;

Figure 13 is a functional block diagram of a decoder in accordance with certain embodiments of the present invention;

Figure 14 is a block diagram of an embodiment of an analog decoder in accordance with the present invention;

Fig. 15 is a block diagram of a component detector according to the embodiment of Fig. 14; and

16 and 17 each show a block diagram of an apparatus according to an embodiment of the present invention having a system for generating audience estimates for widely broadcast information.

Coding

The present invention employs techniques for inserting codes into audio signals to optimize the likelihood that the information in the codes will be accurately retrieved from the signals while ensuring that the codes are inaudible to the human ear when the encoded audio signal is reproduced as sound even if the frequencies of the codes fall within the audible frequency range.

Referring initially to Figure 1, there is illustrated a functional block diagram of an encoder in accordance with one aspect of the present invention.

An audio signal to be encoded is received at an input terminal 30. The audio signal can represent, for example, a program to be broadcast via the radio, the sound range of a television program, a musical composition or any other type of audio signal to be recorded in a certain way. The audio signal can also be a private communication, for example a telephone conversation, or some kind of personal recording. These are of course only examples of the applicability of the present invention and do not represent any intention of restricting the scope of the invention.

As indicated by functional block 34 in Figure 1, the suitability of one or more components of the received audio signal for masking tones having frequencies corresponding to those of the code frequency component or components to be added to the audio signal is assessed . Several evaluations can be carried out for a single code frequency, a separate evaluation can be carried out for each of a number of code frequencies, several evaluations can be carried out for each of a number of code frequencies or one or more joint evaluations for several code frequencies or finally a combination of several of the above options. Each evaluation is made on the basis of the frequency of the code component or components to be masked and the frequency or frequencies of the audio signal component or components whose coverage properties are evaluated. Furthermore, when the code component and the masking audio component or -components do not fall in essentially simultaneous signal intervals, so that they would be reproduced as tones at significantly different time intervals,

AT 41 0 047 B obscures the effects of the differences in signal intervals between the code component or components, and the obscuring program component or components are also considered.

In certain embodiments, multiple evaluations are preferably carried out for each individual code component in that the suitability of different sections of the audio signal for masking each code component is considered separately. In one embodiment, the suitability of each of a number of substantially individual sound audio signal components for masking a code component is assessed on the basis of the frequency of the audio signal component, its amplitude (as defined below) and the timing relevant to the code component, this masking being evaluated below called tonal masking.

The term amplitude is used here for any signal value or for values ​​that can be used to assess the suitability of masking, to select the size of a code component in order to detect its presence in a reproduced signal, or in any other way, including for values ​​such as signal energy , Power, voltage, amperage, intensity and pressure, regardless of whether the measurement takes place on an absolute or relative or on an instantaneous collective value basis. Depending on how it is most favorable, the amplitude can be measured as a window mean value, arithmetic mean value, by integration, as an RMS value (route mean square mean value), as a collective value of absolute or relative discrete values ​​or in any other way.

In other embodiments, in addition to or as an alternative to tonal masking assessments, the suitability of audio signal components within a relatively narrow frequency band that is sufficiently close to a given code component for obscuring the component is assessed (this is hereinafter referred to as narrowband obscuration). In further embodiments, the suitability of multiple code components within a relatively broad frequency band for obscuring the component is assessed. As necessary or appropriate, the suitability of program audio components is assessed at signal intervals preceding or following one or more given components for obscuring those components on a non-simultaneous basis. This type of evaluation is particularly useful when audio signal components in a given signal interval have insufficiently large amplitudes to enable code components with sufficiently large amplitudes to be inserted into the same signal interval so that they can be distinguished from noise.

Preferably, a combination of two or more tonal masking aptitudes, narrowband masking aptitudes, and broadband masking aptitudes (and, if necessary or appropriate, non-simultaneous masking aptitudes) is evaluated for multiple code components. If code components are sufficiently close to one another in terms of their frequency, separate evaluations need not be carried out for each one.

In certain further preferred embodiments, sliding tonal analysis is performed instead of separate tonal, narrowband, and broadband analyzes, eliminating the need to classify the program audio signal as tonal, narrowband, or broadband.

Preferably, when evaluating a combination of masking suitabilities, each evaluation generates a maximum allowable amplitude for one or more code components so that a maximum amplitude can be selected by comparing all evaluations that have been performed and which relate to a given component, which therefore ensures that every component is covered by the audio signal when it is reproduced as sound, so that all components are inaudible to the human ear. Similarly, by maximizing the amplitude of each component, the likelihood that its presence will be detected based on its amplitude is maximized. It is of course not essential that the maximum possible amplitude be used, since in decoding it is only necessary to be able to distinguish a sufficiently large number of code components from audio signal components and other noise.

The results of the evaluations are output, as indicated by 36 in FIG. 1, and made available to a code generator 40. The code generation can be done in many different ways. A particularly advantageous technique has each one of a number

AT 410 047 B assigns a unique set of code frequency components of data states or symbols, so that during a given time interval a corresponding data state is represented by the presence of its corresponding set of code frequency components. In this way, interference with the code detection by audio signal components is reduced, since a sufficiently large number of code components can be detected in an advantageously high percentage of signal intervals despite an interference of the program audio signal with the detection of further components. In addition, the process of implementing the concealment evaluations is simplified if the frequencies of the code components are known before they are generated.

Other forms of coding can also be used. For example, frequency shift keying (FSK), frequency modulation (FM), frequency hopping, coding by fanning out the spectrum and combinations of the aforementioned techniques can be used. Other coding techniques that can be used in practicing the present invention will become apparent from the description herein.

The data to be encoded are received at an input 42 of the code generator 40, which responds to this by generating a unique group of code frequency components and assigning an amplitude to each one on the basis of the evaluations received from the output 36. The code frequency components which are generated in this way are applied to a first input of a summing circuit 46 which receives the audio signal to be coded at a second input. The circuit 46 adds the code frequency components to the audio signal and provides an encoded audio signal at an output terminal 50. The circuit 46 can be either an analog or a digital summing circuit, depending on the form of the signals being fed to the circuit. The summing function can also be provided in software and, if so, a digital processor used to perform the occlusion assessment and to generate the code can also be used to add the code to the audio signal. In another embodiment, the code is provided in the form of time domain data in digital form, which is then added to audio time domain data. In a further embodiment, the audio signal is converted into the frequency domain in digital form and added to the code, which is represented in a similar manner in the form of digital frequency domain data. In most applications, the summed frequency domain data is then converted to time domain data.

It is clear from the following statements that the masking evaluation as well as functions for generating the codes can be carried out either by digital or by analog procedures or by combinations of digital and analog procedures. Furthermore, although the audio signal is received in analog form at the input terminal 30 and in analog form from the circuit 46, as shown in FIG. 1 shown, can be added to the code components, in an alternative embodiment converted to digital form when it is recorded, then added to the code components in digital form and output in either digital or analog form. For example, a signal can be output in digital form if it is to be recorded on a compact disc or on digital audio tape, while it can be output in analog form if it is to be broadcast using conventional radio or television technology. Numerous other combinations of analog and digital approaches can also be used.

In certain embodiments, the code components of only one code symbol are included in the audio signal at a time. In other embodiments, however, the components of multiple code symbols can be contained in the audio signal at the same time. In certain embodiments, for example, the components of a symbol occupy one frequency band and those of another occupy a second frequency band at the same time. In an alternative embodiment, the components of a symbol can be in the same frequency band as another or in an overlapping band, as long as their components are distinguishable, for example by assigning different frequencies or frequency intervals.

One embodiment of a digital encoder is illustrated in FIG. In this embodiment, an audio signal is in analog form at an input terminal 60 and is through

AT 410 047 B converted an A / D converter 62 into digital form. The digital audio signal is provided for masking evaluation, as functionally indicated by block 64, whereupon the digitized audio signal is divided into frequency components, for example by Fast Fourier Transform (FFT), wave transformation (wavelet transform) or another transformation from the time domain to the frequency domain or through digital filtering. Subsequently, the masking suitability of audio signal frequency components within frequency ranges of interest are evaluated with regard to their tonal masking suitability, their narrowband masking suitability and broadband masking suitability (and additionally, if necessary or appropriate, with regard to their non-simultaneous masking suitability). Alternatively, the masking suitability of audio signal frequency components within frequency ranges of interest are assessed with a sliding tonal analysis.

Data to be coded are present at the input terminal 68, the respective group of code components being generated for each data state corresponding to a given signal interval, as indicated by the functional block 72 for signal generation, and being subjected to a level setting, as indicated by the block 76 , which is also supplied with the relevant coverage assessments. The signal generation can also be implemented, for example, by storing each of the code components in a look-up table as time domain data or by interpolating stored data. The code components can either be stored permanently or generated during the initialization of the system according to FIG. The values ​​of the components can also be calculated at the point in time at which they are generated.

The level adjustment is performed for each of the code components on the basis of the relevant masking scores, as explained above, with the code components, the amplitude of which is adjusted to ensure that they are inaudible, added to the digitized audio signal, as indicated by the sum symbol 80 specified. Depending on the length of time that is required to carry out the above-mentioned processes, it can be useful to delay the digital audio signal, as indicated at 82, by temporarily storing it in a memory. If the audio signal is not delayed, the code components adjusted with regard to their amplitude are added after the execution of the FFT and an occlusion evaluation for a first interval of the audio signal to a second interval of the audio signal following the first interval. If, however, the audio signal is delayed, the code components adjusted with regard to their amplitude can instead be added to the first interval, and in this way a simultaneous masking evaluation can be used. If the portion of the audio signal during the first interval has a greater suitability for masking a code component added during the second interval than the portion of the audio signal during the second interval would have with respect to the code component during the same interval, For example, the code component may be assigned an amplitude based on the non-simultaneous masking suitabilities of the portion of the audio signal during the first interval. In this way, both concurrent and non-concurrent masking properties can be evaluated and an optimal amplitude can be assigned to each code component based on the more favorable evaluation.

In certain applications, such as broadcasting or analog recordings (such as on a conventional tape cassette), the encoded audio signal in digital form is converted to analog form by a digital-to-analog converter (DAC) 84. However, if the signal is to be forwarded or recorded in digital form, the DAC 84 can be omitted.

The different functions explained in FIG. 2 can be implemented, for example, by a digital signal processor or by a personal computer, a workstation, a mainframe computer or some other digital computer.

Figure 3 is a block diagram of a coding system for use in coding audio signals that are in analog form, for example in a conventional broadcast studio. In the system according to FIG. 3, a host computer 90, which can for example be a personal computer, monitors the selection and the generation of the information which is to be inserted into a

AT 410 047 B analog audio signal which is present at the input connection 94 is to be coded. The host computer 90 is attached to a keypad 96 and connected to a monitor 100, such as a CRT monitor, so that a user can select a desired message to be encoded, selecting it from a menu of available messages displayed on the Monitor 100 is displayed. A typical message to be encoded in a radio audio signal would be, for example, station or channel identification information, program or segment information and / or a time code.

Once the desired message is entered into the host computer 90, it outputs data representing the symbols of the message to a digital signal processor (DSP) 104, which then receives each symbol it receives from the host computer 90 in the form of a unique set of Code signal components are coded as described below. In one embodiment, the main processor generates a data stream with four states, ie a data stream in which each data unit can assume one of four different data states, each of which represents a unique symbol, including two synchronization symbols, here denoted E and S, and two message information symbols 1 and 0, each of which has a binary state represents. However, any number of different data states can be used and, for example, instead of two message information symbols, three data states can be represented by three unique symbols, which makes it possible to transport a correspondingly larger amount of information with a data stream of a fixed size.

For example, if the program material represents spoken language, it is advantageous to transmit a symbol for a relatively longer period of time than in the case of a program audio signal with a much more continuous energy content to allow for the natural pauses or gaps that exist in spoken language. Accordingly, the number of possible message information symbols is advantageously increased in order to enable the information throughput to be sufficiently large in this case. For symbols representing up to 5 bits, symbol transmission lengths of 2, 3 and 4 seconds create increasingly greater probabilities of correct decoding. In some such embodiments, an initial symbol (E) is decoded when (i) the energy in the FFT regions for that symbol is greatest, (ii) the average energy minus the standard deviation of the energy for that symbol is greater than the average energy plus the average standard deviation of energy for all other symbols, and (iii) the shape of the time course of energy for that symbol is generally bell-shaped and peaks at the temporal boundary between symbols.

In the embodiment of FIG. 3, when the DSP 104 has received the symbols of a given message to be encoded, it responds by generating a unique set of code frequency components for each symbol which it provides at the output 106. For each of the four data symbols S, E, 0 and 1 of the exemplary data set described above, spectral diagrams are given with reference to FIG. 4. As shown in Fig. 4th As can be seen, in this embodiment the symbol S is represented by a unique group of 10 code frequency components ft to f₁₀ shown, which are arranged at equal frequency intervals in a range that extends from a frequency value of slightly above 2 kHz to a frequency value slightly below 3 kHz. The symbol E is represented by a second, unambiguous group of 10 code frequency components f "to f₂o, which are arranged in the frequency spectrum at equal intervals from a first frequency value of slightly above 2 kHz to a frequency value of slightly less than 3 kHz, each of the code components f₁₄ to f₂₀ has a unique frequency value that is different from all others in the same group as well as from all frequencies ft through fio. The symbol 0 is represented by a further, unambiguous group of 10 code frequency components f₂₁ to f₃₀ which are also arranged with equal frequency intervals from a value slightly above 2 kHz to a value of slightly less than 3 kHz and each of which has a unique frequency value that differs from all the others in the same group as well as from all frequencies L to f₂₀ differs. Finally, the symbol 1 is represented by a further, unique group of 10 code frequency components f₃₁ to f₄₀shown, which are also arranged at equal frequency intervals from a value of a little more than 2 kHz to a value of a little less than 3 kHz, so that each of the Kompo11

AT 410 047 B nenten f₃₁ to f₄₀ has a unique frequency value that differs from each of the other frequency components f-ι to f₄₀ differs. By using multiple code frequency components for each data state so that the code components of each state are substantially separated in frequency, the presence of noise (such as audio signal components other than coding, or other noise) in a common detection band with another code component of a given data state interferes with the detection of the other components of this data state.

In other embodiments it is advantageous to represent the symbols by multiple frequency components, for example 10 code tones or frequency components which are not uniformly spaced in frequency and which do not have the same offset from symbol to symbol. If an integral relationship between code frequencies for a symbol is avoided by grouping the tones, the effects of intermediate frequency superimpositions and space cancellations, ie of places where echoes from room walls interfere with correct decoding.

The following sets of code tone frequency components for the four symbols (0, 1, S and E) are given to reduce the effects of space cancellations, where ft through f₁₀ the respective code frequency components of each of the four symbols mean (expressed in Hertz):

<td></td><td> 0</td><td> 1</td><td>S.</td><td>Έ</td>

<td>f1</td><td> 1046,9</td><td> 1054,7</td><td> 1062,5</td><td> 1070,3</td>

<td>f2</td><td> 1195,3</td><td> 1203,1</td><td> 1179,7</td><td> 1187,5</td>

<td>f3</td><td> 1351,6</td><td> 1343,8</td><td> 1335,9</td><td> 1328,1</td>

<td>f4</td><td> 1492,2</td><td> 1484,4</td><td> 1507,8</td><td> 1500,0</td>

<td>f5</td><td> 1656,3</td><td> 1664,1</td><td> 1671,9</td><td> 1679,7</td>

<td>f6</td><td> 1859,4</td><td> 1867,2</td><td> 1843,8</td><td> 1851,6</td>

<td>f7</td><td> 2078,1</td><td> 2070,3</td><td> 2062,5</td><td> 2054,7</td>

<td>f8</td><td> 2296,9</td><td> 2289,1</td><td> 2304,7</td><td> 2312,5</td>

<td>f9</td><td> 2546,9</td><td> 2554,7</td><td> 2562,5</td><td> 2570,3</td>

<td>f10</td><td> 2859,4</td><td> 2867,2</td><td> 2843,8</td><td> 2851,6</td>

Generally speaking, in the examples given above, the spectral content of the code varies relatively little when the DSP 104 switches its output from one of the data states S, E, 0 and 1 to another of these. In accordance with one aspect of the present invention, in certain preferred embodiments, each code frequency component of each symbol is paired with a frequency component of every other data state so that the difference between them is less than the critical bandwidth therefor. For any pair of pure tones, the critical bandwidth is a frequency range within which the frequency difference between the two tones can be changed without significantly increasing the loudness. Since the frequency spacing between adjacent tones is the same in the case of each data state S, E, 0 and 1, and since each tone of each data state S, E, 0 and 1 is paired with a tone of each other, so that the difference the frequency in between is less than the critical bandwidth for that pair, there is essentially no change in loudness in going from any of the data states S, E, 0 and 1 to any other of them, when reproduced as sound. Moreover, by minimizing the frequency difference between the code components of each pair, the relative likelihood of each data condition being detected when it is received is not significantly affected by the frequency characteristics of the transmission path. Another benefit of pairing components of different data states so that they are relatively close in frequency is that an occlusion assessment performed on a code component of a first data state is substantially accurate for a corresponding component of a next data state if a switching of the states takes place.

Alternatively, the non-uniform coded tone spacing scheme is used to minimize the

AT 410 047 B

The effects of the space cancellations clearly show that the frequency components ff to f for each of the code₁₀ selected frequencies are grouped around a frequency, for example the frequency components for ff, f₂and f₃ arranged in the vicinity of 1,055 Hz, 1,180 Hz and 1,340 Hz. In particular, in this exemplary embodiment, the tones have a mutual spacing of twice the FFT resolution, for example, for a resolution of 4 Hz, the tones are shown at a mutual spacing of 8 Hz and are selected so that they are in the middle of the frequency range an FFT area. Furthermore, the order of the different frequencies that correspond to the code frequency components ff to ff₀ to represent the different symbols 0, 1, S and E are assigned, changed in each grouping. For example, the frequencies that correspond to the components ff, f₂ and f₃ are selected, the symbols (0,1, S, E), (S, E, 0, 1) or (E, S, 1, 0) from the lowest to the highest frequency, i.e. (1046.9, 1054, 7, 1062.5, 1070.3), (1179.7, 1187.5, 1195.3, 1203.1), (1328.1, 1335.9, 1343.8, 1351.6). An advantage of this arrangement is that even if space cancellation occurs which interferes with the correct reception of a code component, generally the same tone is eliminated from each of the symbols so that it is easier to assign a symbol from the remaining components decode. In contrast, when space cancellation eliminates a component from one symbol but not from another symbol, it is more difficult to correctly decode the symbol.

In an alternative embodiment, either more or less than four separate data states or symbols can be used for the coding. Also, each data state or symbol can be represented by more or less than 10 code tones, and although it is preferred that the same number of tones be used to represent each data state, it is not essential in every application that the number of code tones used to represent each data state is the same. Preferably, each of the code tones differs in frequency from all other code tones in order to maximize the likelihood that each of the data states can be distinguished during decoding. However, it is not essential in all applications that none of the frequencies of the code tones are the same for two or more data states.

FIG. 5 is a functional block diagram to which reference will be made in explaining the coding function performed in the embodiment of FIG. As mentioned above, the DSP 104 receives data from the main processor 90 which designate the sequence of data states which are to be output by the DSP 104 in each case as groups of code frequency components. Preferably, the DSP 104 creates a look-up table of time domain representations for each of the code frequency components ff through f₄₀which it then stores in its RAM represented by memory 110 in FIG. In response to the data received from host computer 90, DSP 104 generates a respective address which it applies to an address input from memory 110, designated 112 in FIG. 5, to cause memory 110 to output time domain data for each of the 10 frequency components that correspond to the data status to be output at this point in time.

Still referring to FIG. 6, which is a functional block diagram illustrating certain operations performed by DSP 104, memory 110 stores a sequence of time domain values ​​for each of the frequency components of each symbol S, E, 0 and 1. In this particular embodiment, since the code frequency components range from about 2 kHz to about 3 kHz, a sufficiently large number of time domain signals are used for each of the frequency components ff to f₄₀ are stored in memory 110 so that they can be output at a rate greater than the Nyquist frequency of the code component having the highest frequency. The time domain code components are output at a correspondingly high speed from the memory 110 which stores the time domain components for each of the code frequency components representing a predetermined period of time, so that (n) time domain components for each of the code frequency components ff to f₄₀ for (n) time intervals ff to t "are stored, as FIG. 6 shows. For example, if the symbol S is to be encoded during a given signal interval during the first interval ff, the memory 110 outputs the time domain components ff to f₁₀which correspond to this interval, as they are stored in the memory 110. During the next interval the time domain components ff to f₁₀ for the interval t₂ output from memory 110. This

AT 410 047 B

Process continues sequentially for the intervals t₃ are you_n and back to t-ι continued until the duration of the coded symbol S has expired.

In certain embodiments, instead of outputting all 10 code components, ie for example T to f₁₀, only those code components are output during a time interval that lie within the critical bandwidth of the tones of the audio signal. This is a generally conservative approach to ensure the inaudibility of the code components.

Reference is again made to FIG. 5. The DSP 104 also serves to adjust the amplitudes of the time domain components output from the memory 110 so that the code frequency components, when reproduced as sound, are masked by the components of the audio signal in which they are inserted, so that they are inaudible to the human ear. As a result, the DSP 104 is also supplied with the audio signal present at the input terminal 94, after this has been suitably filtered and converted from analog to digital form. In particular, the encoder according to FIG. 3rd an analog band filter 120, which serves to remove audio signal frequency components substantially which are outside a band which is of interest for the evaluation of the suitability of masking of the recorded audio signal and which in the present embodiment is approximately 1.5 kHz extends to about 3.2 kHz. The filter 120 is also used to remove high frequency components from the audio signal that are band-overlapping or Cause aliasing effects; when the signal is then digitized by an analog-to-digital converter (A / D) 124 which operates at a sufficiently large sampling frequency.

As can be seen from FIG. 3, the digitized audio signal is passed through the A / D converter 124 to the DSP 104, where the program audio signal, as denoted by 130 in FIG. 5, is subjected to frequency domain separation. In this special embodiment, the frequency range separation is carried out as a fast Fourier transform (FFT), which is carried out periodically with or without temporary overlap in order to generate successive frequency ranges which each have a predetermined frequency width. Further techniques are available for separating the frequency components of the audio signal, for example the wave transformation (wavelet transform), the discrete Walsh-Hadamard transformation, the discrete Hadamard transformation, the discrete cosine transformation and various digital filtering techniques.

When the DSP 104 has divided the frequency components of the digitized audio signal into the successive frequency ranges, as explained above, it comes to assess the suitability of different frequency components present in the audio signal with regard to the concealment of the different code components which are output from the memory 110, and to the generation respective amplitude adjustment factors, which are used to to divide the amplitudes of the different code frequency components so that they are covered by the program audio signal when they are reproduced as sound, so that they are inaudible to the human ear. These processes are represented in FIG. 5 by block 134.

For audio signal components that are substantially simultaneous with the code frequency components they are intended to obscure (but precede the code frequency components by a short period of time), the obfuscation suitability of the program audio components is assessed on a tonal basis, as well as on a narrowband basis. Cloaking basis and on a broadband cloaking basis as described below. For each code frequency component output from memory 110 at a given point in time, a tonal masking suitability for each of a number of audio signal frequency components is assessed based on the energy level in each of the respective ranges within which those components fall also based on the frequency relationship of each area to the respective code frequency component. The evaluation can in any case (tonal, narrowband and broadband) take the form of an amplitude adjustment factor or some other measure, whereby it is possible to assign an amplitude to the code component so that the code component is covered by the audio signal. Alternatively, the evaluation can be a sliding tonal analysis.

In the case of narrow band masking, in this embodiment the energy content of frequency components is below one for each relevant code frequency component

AT 410 047 B rated levels within a predetermined frequency band including the corresponding code frequency component in order to derive a separate assessment of the suitability for masking. In certain implementations, the possibility of narrow band masking is measured based on the energy content of those audio signal frequency components that are below the average energy level of the region within the predetermined frequency band. In this embodiment, the energy levels of the components below the energy levels of the components that are below the average energy of the region (as the component threshold) are summed up in order to generate a narrowband energy level, in response to which a corresponding narrowband masking rating is identified for the respective code component. Instead, a different narrowband energy level can be generated by selecting a different component threshold than the average energy level. In other embodiments, the average energy level of all audio signal components within the predetermined frequency band is used as the narrowband energy level in order to assign a narrowband occlusion rating to the respective code component. In still other embodiments, instead, the entire energy content of the audio signal components within the predetermined frequency band is used, while in other embodiments a minimum component level within the predetermined frequency band is used for this purpose.

Finally, in certain embodiments, the broadband energy content of the audio signal is determined in order to assess the suitability of the audio signal for masking the respective code frequency component on a broadband masking basis. In this embodiment, the broadband occultation score is based on the minimum narrowband energy level found in the course of the narrowband occlusion scores described above. That is, if four separate predetermined frequency bands are examined in the course of the narrowband occlusion evaluation described above, and if broadband noise includes the minimum narrowband energy level among all four predetermined frequency bands (whichever is determined), then this minimum narrowband energy level becomes with multiplied by a factor, which is equal to the ratio of the frequency range spanned by all four narrow bands to the bandwidth of the predetermined frequency band with the minimum narrow band energy level. The resulting product indicates an allowable overall code performance level. If this total allowable code power level is denoted by P and the code includes ten code components, an amplitude adjustment factor is assigned to each to achieve a component power level that is 10 dB less than P. In an alternative embodiment, the broadband noise for a predetermined, relatively wide band comprising the code components is calculated by selecting one of the techniques discussed above for estimating the narrowband energy level, but instead using the audio signal components across the predetermined, relatively wide band . When the broadband noise is determined in the chosen manner, each code component is assigned a corresponding broadband masking score.

The amplitude adjustment factor for each code frequency component is then selected on the basis that the tonal, narrowband or broadband masking evaluation results in the highest permissible level of the respective component. This maximizes the likelihood that each code frequency component can be distinguished from non-audio noise, while at the same time ensuring that the respective code frequency component is obscured so that it cannot be perceived by the human ear.

The amplitude adjustment factors are selected for each of the tonal, narrow band and wide band masking based on the following factors and circumstances. In the case of tonal masking, the factors are assigned on the basis of the frequencies of the audio signal components whose masking suitability is being evaluated and the frequency or frequencies of the code components to be masked. Furthermore, a given audio signal over any selected interval provides the ability to obscure a given code component within the same interval (i.e., simultaneous obscuring) to a maximum level greater than that which the same audio signal is capable of over the selected interval , the same code component before or after the selected interval

AT 410 047 Β occurs (ie not concealing simultaneously). The conditions under which the encoded audio signal is perceived by an audience or another audience group, depending on the case, are preferably also taken into consideration. If, for example, the audio signal of a television broadcast is to be encoded, the distortion effects of a typical listening environment are preferably taken into account, since in such environments certain frequencies are attenuated more strongly than others. Recording and playback devices (such as graphic equalizers) can cause similar effects. Effects associated with the environment and with the technical equipment can be compensated for by choosing setting factors with sufficiently small amplitudes to ensure that masking takes place under the expected conditions.

In certain embodiments, only either the tonal or the narrow-band or the broad-band masking suitability is assessed. In other embodiments, two of these different types of masking suitability are assessed, while in still other embodiments all three are used.

In certain embodiments, sliding tonal analysis is used to assess the suitability of the audio signal for masking. A sliding tonal analysis generally satisfies the masking criteria for narrowband noise, broadband noise and individual tones without a classification of the audio signal being necessary. In sliding tonal analysis, the audio signal is viewed as a set of discrete tones, each centered on a range of FFT frequencies. In general, the sliding tonal analysis first calculates the power of the audio signal in each FFT range. Then, for each coding tone, the masking effects of the discrete tones of the audio signal in each FFT area which is separated in frequency from such coding tones by no more than the critical bandwidth of the audio tone are weighted on the basis of the audio signal power in each such area, the Masking relationships can be used for single tone masking. The masking effects of all relevant discrete tones of the audio signal are summed up for each coding tone and then adjusted according to the number of tones within the critical bandwidth of the audio signal tones and the complexity of the audio signal. As will be described below, the complexity of the program material in certain embodiments is empirically based on the ratio of the power in the relevant tones of the audio signal and the square root of the sum of the squares of the powers in these audio signal tones. The complexity serves to take into account the fact that the narrow-band noise and the broad-band noise each have much better masking effects than can be obtained from a simple addition of the tones that are used to map the narrow-band and broad-band noise.

In certain embodiments in which a sliding tonal analysis is used, a predetermined number of samples of the audio signal are first subjected to a large FFT which brings a high resolution but requires a longer processing time. Subsequently, successive sections of the predetermined number of samples are subjected to a relatively less extensive FFT, which is faster but brings a lower resolution. which generally corresponds to a time weighting of the extensive FFT with a higher frequency accuracy due to the greater time accuracy of the less extensive FFT.

In the embodiment of FIG. 5, the DSP 104 adjusts the amplitude of each code frequency component accordingly, as indicated by the functional block Amplitude Adjustment 114, once a suitable amplitude adjustment factor has been selected for each of the code frequency components output from the memory 110 . In other embodiments, each code frequency component is initially generated so that its amplitude corresponds to the respective adjustment factor. Referring also to FIG. 6, in the amplitude adjusting process of the DSP 104 in this embodiment, the ten selected from the values ​​fi to f₄₀ the code frequency components in the time domain for the current time interval ti to t_n are selected, with a respective amplitude adjustment factor G_A1 to G_A10 multiplied, and then the DSP 104 adds the amplitude adjusted time domain components to generate a composite code signal that is provided at its output 106

AT 410 047 B becomes. The composite code signal, with reference to FIGS. 3 and 5, is converted into analog form by a digital / analog converter (DAG) 140 and given to a first input of a summing circuit 142. The summing circuit 142 receives the audio signal from the input terminal 94 at a second input and adds the composite, analog code signal to the analog audio signal and outputs an encoded audio signal to an output 146.

In (radio) broadcast applications, the encoded audio signal modulates a carrier wave and is then emitted. In NTSC television broadcasting applications, the encoded audio signal frequency modulates a subcarrier and is mixed with a composite video signal so that the combined signal is used to modulate a broadcast carrier for wireless broadcasting. These radio and television signals can of course also be transmitted via cable (e.g. via conventional or optical fiber cables), via satellites or in any other way. In other applications, the encoded audio signal can be recorded either for distribution in recorded form or for subsequent broadcast or other redistribution. Encoded audio signals can also be used for point-to-point transmissions. There are numerous other applications as well as transmission and recording techniques that can be used here.

7A to 7C show flow charts for explaining a software program executed by the DSP 104 to implement the evaluation of the tonal, narrowband and broadband evaluation functions described above. Fig. 7A shows a main loop of the software program of the DSP 104. The program is executed by an instruction from the main or Central computer 90 initiates (step 150) whereupon the DSP 104 initializes its hardware registers (step 152) and then goes to step 154 ​​to calculate unweighted code component data in the time domain, as shown in FIG. 6, which is then stored in memory, to be read out as needed to generate the code components in the time domain, as explained above. In an alternative embodiment, this step can be omitted if the code components are stored permanently in a ROM or some other non-volatile memory. It is also possible to calculate the code component data if necessary, although this increases the required computing power. Another alternative is to generate unweighted code components in analog form and then to adjust the amplitudes of the analog components by means of weighting factors that are generated by a digital computer.

When the time domain data is calculated and stored, the DSP 104 transmits a request to the host computer 90 in step 156 to encode a next message. The message consists of a chain of characters, integer numbers or some other set of data symbols which uniquely identify the code component groups to be output by the DSP 104 in a sequence predetermined by the message. In another embodiment, the main computer itself determines, knowing the output data rate of the DSP, when a next message is to be sent to the DSP by setting a suitable timer and issuing the message as soon as a state occurs which corresponds to a timeout. In a further alternative embodiment, a decoder is coupled to the output of the DSP 104 to receive the code components issued, to decode them and to return the message to the main processor as it is issued by the DSP so that the main computer can determine when another one Message to the DSP 104 is to be given. In still other embodiments, the functions of main processor 90 and DSP 104 are performed by a single processor.

As soon as the next message has been received by the host computer, following step 156, the DSP proceeds to generate the code components for each symbol of the message in the correct order and to provide the combined, weighted code frequency components at its output 106.

This process is represented by a loop labeled 160 in FIG. 7A.

After entering the loop labeled 160, the DSP 104 activates timer interrupts 1 and 2 and then enters a subroutine for calculating the weighting factors described in connection with the flowcharts of Figures 7B and 7C. Referring first to Figure 7B. After entering the subroutine 162 for calculating the

AT 410 047 B

Weighting factors, the DSP first determines whether a sufficient number of samples of the audio signal are stored to enable a high-resolution FFT to be performed to analyze the spectral content of the audio signal during a predetermined audio signal interval that is at least long ago, as in step 163 designated. When starting the method, a sufficient number of samples of the audio signal must first be accumulated in order to be able to carry out the FFT. However, if an overlapping FFT is used, fewer data values ​​need to be stored during successive loops before the next FFT is executed.

As can be seen from Figure 7B, the DSP loops at step 163 and waits for the necessary accumulation of samples. At each timer interrupt 1, the A / D 124 provides a new digitized sample of the program audio signal that is accumulated in a data buffer of the DSP 104, as indicated by the subroutine 164 in FIG. 7A.

Once a sufficiently large number of data values ​​has been accumulated by the DSP, referring again to FIG. 7B, processing continues to step 168 where the aforementioned high resolution FFT is applied to the audio signal data samples from the least recent audio signal interval is applied. Thereafter, a respective weighting factor or amplitude adjustment factor for each of the ten code frequency components in the symbol that is being coded is calculated, as indicated by 170. In a further step 172 that of the frequency ranges generated by the high-resolution FFT (step 168) is determined in the manner explained above which is suitable for covering the highest level of the respective code component on a single tone basis (the tonal dominant).

Referring to Figure 7C, in a step 176 the weighting factor for the tonal dominant is determined and stored for comparison with relative masking properties offered by the narrowband and broadband masking and is used when it is found to be the most effective Masking is used as a weighting factor to adjust the amplitude of the current code frequency component. In a subsequent step 180, an assessment of the narrow-band and broad-band masking suitability is carried out in the manner described above. Thereafter, in a step 182 it is determined whether the narrow-band masking has the best suitability for masking the respective code component, and if so, the weighting factor is provided with an update in a step 184 on the basis of the narrow-band masking. In a subsequent step 186 it is determined whether the broadband masking is best suited for masking the respective code frequency component, and if so, in a step 190 the weighting factor for the respective code frequency component is set on the basis of the broadband masking. Next, at step 192, it is determined whether weighting factors have been selected for each one of the code frequency components currently to be output to represent the current symbol, and if not, the loop is re-initiated to select a weighting factor for the next code frequency component. If, however, the weighting factors for all of the components have been selected, the subroutine terminates as step 194 indicates.

When a timer interrupt 2 occurs, the sequence goes to a subroutine 200 in which the functions explained in FIG. 6 are carried out. This means that in subroutine 200 the weighting factors which were calculated in the course of subroutine 162 are used to multiply the respective time domain values ​​of the current symbol to be output, the weighted code component values ​​of the time domain being added and sent to the DAC as a weighted, composite code signal 140 can be issued. Each code symbol is output for a predetermined period of time, after which the process returns from step 202 to step 156.

Figures 7D and 7E are flow charts illustrating an implementation of the sliding tonal analysis technique for evaluating the masking effects of an audio signal. In step 702, variables are initialized, for example the size with regard to the data values ​​of a large FFT and a less extensive FFT, the number of less extensive FFTs in a large FFT and the number of code tones per symbol, for example 2048, 256, 8 and 10.

AT 410 047 B

In steps 704 through 708, a number of samples are analyzed that corresponds to a large FFT. At step 704 the samples of the audio signal are obtained. At step 706, the performance in the program material in each FFT area is obtained. In step 708, the allowable code tone power in each corresponding FFT range is obtained for each of the tones, taking into account the effects of all relevant audio signal tones within that range. The flow chart of FIG. 7E shows step 708 in greater detail.

In steps 710-712, a number of samples corresponding to a less extensive FFT are analyzed, similar to steps 706-708 in the case of a large FFT. In step 714, the permissible code powers which have been found on the basis of the extensive FFT in step 708 and the less extensive FFT in step 712 are combined for the part of the samples that were subjected to a less extensive FFT. At step 716 the code tones are mixed with the audio signal to produce an encoded audio signal, and at step 718 the encoded audio signal is passed to the DAC 140. At step 720 it is decided whether steps 710 to 718 are to be repeated, ie whether there are parts of audio samples which have been subjected to an extensive FFT but not to a less extensive FFT. Then, in step 722, a next number of sample values ​​is analyzed in accordance with a comprehensive FFT, provided that further audio signal sample values ​​are present.

Fig. 7E shows details for steps 708 and 712 where the allowable code power in each FFT area is calculated. In general, this approach maps the audio signal to include a set of tones (see example below), calculates the masking effect of each audio tone for each code, sums the masking effects and makes adjustments to the density of the code tones and the complexity of the audio signal .

At step 752 the band of interest is determined. The bandwidth used for coding is, for example, 800 Hz to 3200 Hz and the sampling frequency 44100 values ​​/ second. The start range starts at 800 Hz and the end range ends at 3200 Hz.

At step 754, the masking effect of each relevant audio tone for each code in that area is determined by using the masking curve for a single tone and by determining the non-zero width of the FFT range of the audio signal by determining (1) a first masking value based on the assumption that the total audio signal power is at the high end of the range, and (2) a second masking value based on the assumption, that the total audio signal power is at the lower end of the range, and then selecting the smaller value of said first and second masking values ​​is compensated.

Fig. 7F shows an approximation of a single-tone masking curve for an audio signal tone with a frequency of fPGM, which in this example is about 2200 Hz, according to Zwislocki, J: J., Masking: Experimental and Theoretical Aspects of Simultaneous, Forward, Backward and Central Masking ( Covering: Experimental and theoretical aspects of simultaneous, forward, backward and central covering), 1978, in Zwicker et al., Ed., Psychoacoustics: Facts and Models, pages 283-316, Springer-Verlag, New York. The width of the critical band (CB) is defined by Zwislocki as follows:

Critical band = 0.002 * f_PGM^1,5 +100

With the following definitions, whereby masking of the audio signal tone is +/- 0.3 critical bands /

-16 dB from the cover / -26 dB from the cover / -24 dB per critical band /

-7 dB per critical band /

BRKPOINT = 0.3 /

PEAKFAC = 0.025119 /

BEATFAC = 0.002512 / mNEG = 2.40 / mPOS = -0.70 / cf = code frequency mf = masking frequency cband = critical band around f_PGM

AT 410 047 B the masking factor, mFactor, can be calculated as follows:

brkpt = cband + BRKPOINT if on the negative slope of the curve according to FIG. 7F, mfactor = PEAKFAC * 10 ** (mNEG * mf-brkpt-cf) / cband) if on the flat part of the curve according to FIG. 7F, mfactor = BEATFAC if on the positive slope of the curve according to FIG. 7F, mfactor = PEAKFAC * 10 ** (mPOS * cf-brkpt-mf) / cband).

Specifically, a first mfactor is calculated based on the assumption that the total audio signal power is at the low end of its range, then a second mfactor is calculated based on the assumption that the total audio signal power is at the high end of its range and it becomes the smaller of these two m-factor values ​​is selected as the masking value generated by this audio signal tone for the selected code tone. In step 754, this process is performed for each relevant audio tone for each codone.

In step 756, each code tone is adjusted by each of the masking factors corresponding to the audio signal tones.

In this embodiment, the masking factor is multiplied by the audio signal power in the relevant area.

In step 758, the result of the multiplication of the concealment factors by the audio signal power is summed up for each area in order to obtain an allowable power for each code tone.

In step 760 the permissible code tone powers are set with regard to the code tones within a critical bandwidth on both sides of the code tone which is currently being evaluated and with regard to the complexity of the audio signal. The number of code tones within the critical band, CTSUM, is counted. The adjustment factor, ADJFAC, results from:

ADJFAC = GLOBAL * (PSUM / PRSS)^1,5 / CTSUM where GLOBAL is an attenuation factor that takes into account the coding inaccuracy due to time delays in the FFT execution, (PSUM / PRSS)^1,5 is an empirical correction factor for the complexity, and 1 / CTSUM simply means that the audio signal power is divided among all the code tones it is intended to obscure. PSUM is the sum of the masking tone power levels assigned to masking the codetone whose ADJFAC is being determined. The square root of the sum of the squares of the powers (PRSS) results from

PRSS = SQRT (Σ (P²), i - FFT areas in the band

I.

Assuming that a total existing masking tone power in a band is evenly distributed over, for example, one, two and three tones, the following results:

AT 410 047 B

<td>number Sounds</td><td>Sound performance</td><td>PSUM</td><td>PRSS</td>

<td> 1</td><td> 10</td><td> 1 * 10= 10</td><td> 10</td>

<td> 2</td><td> 5,5</td><td> 2*5 = 10</td><td>SQRT (2.5²) = 7,07</td>

<td> 3</td><td> 3,3, 3,3, 3,3</td><td> 3*3,3=10</td><td>SQRT (3 * 3.3²) = 5,77</td>

PRSS is therefore a measure of whether the concealment performance is more in the form of a peak value (increasing values) or in the form of a broader distribution (decreasing values) of the program material.

Step 762 in Fig. 7E determines whether there are any more areas in the band of interest, and if so, the procedure described above is followed.

Some examples of occlusion calculations are given below. An audio signal symbol is assumed at 0 dB, so that the values ​​provided are the maximum code tone powers with respect to the audio signal power. There are four cases: a single tone at 2500 Hz; three tones at 2000, 2500 and 3000 Hz; narrow band noise, modeled as 75 tones within the critical band and centered at 2600 Hz, ie 75 Tones evenly spaced 5 Hz in the range of 2415 to 2785 Hz; and broadband noise, modeled as 351 tones evenly spaced 5 Hz in the range of 1750 Hz to 3250 Hz. For each case, a result calculated by means of a sliding tonal analysis (STA) is compared with a calculated result obtained from a selection of the best analysis among the single-tone, narrow-band noise and broad-band noise analysis.

<td></td><td colspan="2">Single tone</td><td colspan="2">Multiple tones</td><td colspan="2">Narrow band rush</td><td colspan="2">Broadband rush</td>

<td>Codetone (Hz)</td><td>STA (dB)</td><td>Best of 3 (dB)</td><td>STA (dB)</td><td>Best of 3 (dB)</td><td>STA (dB)</td><td>Best of 3 (dB)</td><td>STA (dB)</td><td>Best of 3 (dB)</td>

<td> 1976</td><td> -50</td><td> -49</td><td> -28</td><td> -30</td><td> -19</td><td>N / A</td><td> 14</td><td> 12</td>

<td> 2070</td><td> -45</td><td> -45</td><td> -22</td><td> -32</td><td> -14</td><td>N / A</td><td> 13</td><td> 12</td>

<td> 2163</td><td> -40</td><td> -39</td><td> -29</td><td> -25</td><td> -9</td><td>N / A</td><td> 13</td><td> 12</td>

<td> 2257</td><td> -34</td><td> -33</td><td> -28</td><td> -28</td><td> -3</td><td>N / A</td><td> 12</td><td> 12</td>

<td> 2351</td><td> -28</td><td> -27</td><td> -20</td><td> -28</td><td> 1</td><td>N / A</td><td> 12</td><td> 12</td>

<td> 2444</td><td> -34</td><td> -34</td><td> -23</td><td> -33</td><td> 2</td><td> 7</td><td> 13</td><td> 12</td>

<td> 2538</td><td> -34</td><td> -34</td><td> -24</td><td> -34</td><td> 3</td><td> 7</td><td> 13</td><td> 12</td>

<td> 2632</td><td> -24</td><td> -24</td><td> -18</td><td> -24</td><td> 5</td><td> 7</td><td> 14</td><td> 12</td>

<td> 2726</td><td> -26</td><td> -26</td><td> -21</td><td> -26</td><td> 5</td><td> 7</td><td> 14</td><td> 12</td>

<td> 2819</td><td> -27</td><td> -27</td><td> -22</td><td> -27</td><td> 6</td><td>N / A</td><td> 15</td><td> 12</td>

For example, in the sliding tonal analysis (STA), the masking tone for the single tone is 2500 Hz, which is a critical bandwidth of 0.002 * 2500^1,5 + 100 = 350 Hz. The transition points of the curve in FIG. 7F are at 2500 ± 0.3 * 350 or 2395 and 2605 Hz. The code frequency from 1976 is evident on the section with a negative slope of the curve according to FIG. 7F, so that the masking factor becomes:

mfactor = 0.025119 * 10 '²'⁴ * (2500-105-1976)/350 = 3,364 *10’⁵= -44.7 dB

Since there are three code tones within the critical band of 1976 Hz, the concealment power is divided between them:

AT 410 047 B

3,364 * 10'⁵ / 3 = -49.5 dB

This result is rounded to the -50 dB shown in the upper left part of the sample calculation table.

In the best of three analysis, tonal occlusion is calculated according to the single tone method discussed above in connection with Figure 7F.

In the best of 3 analysis, narrowband noise masking is calculated by first calculating the mean power over a critical band centered on the frequency of the codetone of interest. Sounds with a power greater than the mean power are not considered part of the noise and are removed. The sum of the remaining power is the narrowband noise power. The maximum permissible code tone power is -6 dB of the narrowband noise power for all code tones within a critical bandwidth of the codetone of interest.

In the best of 3 analysis, broadband noise masking is calculated by calculating the narrowband noise power for critical bands centered at 2000 Hz, 2280 Hz, 2600 Hz and 2970 Hz. The smallest resulting narrowband noise power is multiplied by the ratio of the total bandwidth to the corresponding critical bandwidth to find the broadband noise power. For example, if the band centered at 2600 Hz with a critical bandwidth of 370 Hz is the minimum, its narrow band noise power is multiplied by 1322 Hz / 370 Hz = 3.57 to get the broad band noise power. The permissible coded tone power is -3 dB of the broadband noise power. If there are ten code tones, the maximum allowable power for each is 10 dB less or -13 dB of the broadband noise power.

It can be seen that the calculations after the moving tonal analysis generally correspond to the best of 3 calculations, indicating that the moving tonal analysis is a robust method. In addition, the results obtained by the sliding tonal analysis in the case of multiple tones are better, ie they allow for greater codetone powers than the best of three analysis, indicating that the moving tonal analysis is suitable even for cases that do not exactly fit into one of the best of three calculations.

Referring now to Figure 8, there is shown in block form one embodiment of an encoder employing analog circuitry. The analog encoder receives an audio signal in analog form at an input terminal 210, from which the audio signal is input to N component generation circuits 220-i to 200_N is passed, each of which has a code component Ci to C_N generated. For the sake of simplicity and clarity of illustration, only the component generation circuits 220i and 220 are shown_N shown in FIG. In order to generate the code components of a respective data symbol which are to be inserted into the audio signal for generating an encoded audio signal in a controllable manner, each component generation circuit is provided with a respective data input connection 222! to 222_n provided, which serves as an activation input for the respective component generation circuit. Each symbol is created as a subset of the code components Ci to C._N encoded by sending an activation signal selectable to specific, individual component generation circuits 220i to 220_n is created. The generated code components corresponding to each data symbol are fed as inputs to a summing circuit 226 which receives the input audio signal from input terminal 210 at a further input and which serves to add the code components to the input audio signal in order to generate the encoded audio signal, that she delivers at one of her exits.

The individual component generation circuits are constructed in a manner similar to one another and each contain a weighting factor determination circuit 230! to 230_N, a respective signal generator 232_Ί to 232_N and switching circuits 234i to 234, respectively_N. Each of the signal generators 232_Ί to 232_N generates a respectively different code component frequency and forwards the generated component to the respective switching circuit 234_Ί to 234_N, each of which has a second input connected to ground and an output which is connected to an input of a respective multiplication circuit 236i to 236_N connected. In response to receiving an activation input on their respective data input ports 222i-222_n couples each of the switching circuits 234i to 234_N the outcome of his respective

AT 410 047 B

Signal generator 232₄ to 232_N to the input of the respectively corresponding multiplication circuit 236i to 236_N. In the absence of an activation signal at the data input, however, each switching circuit 234 ·] to 234 couples_N its output to the grounded input, so that the output of the corresponding multiplier 236i to 236_N is on 0.

Each face factor determination circuit 230i to 230_N is used to assess the suitability of frequency components of the audio signal within a corresponding frequency band thereof with regard to the coverage of the code component which is generated by the corresponding signal generator 232! to 232_N has been generated to produce a weighting factor which it is input to the corresponding multiplying circuit 236i to 236_N passes on in order to adjust the amplitude of the corresponding code component, so that it is ensured that this is covered by the portion of the audio signal which has been weighted by the weighting factor determination circuit. Still referring to FIG. 9, the construction of each weight factor determination circuit 230 is i to 230_N explained in block form, with an exemplary circuit 230 is given. The circuit 230 includes a masking filter 240 which receives an audio signal at an input thereof and is used to filter the portion of the audio signal which is used to generate a weighting factor which is used for the respective multiplier 236! to 236_N is conducted, is used to separate. The properties of the masking filter are further selected so that the amplitudes of the audio signal frequency components are weighted according to their relative suitability for masking the respective code component.

The portion of the audio signal selected by the masking filter 240 is applied to an absolute value circuit 242 which generates an output signal representing an absolute value of a portion of the signal within the frequency band passed through the masking filter 240. The output of the absolute value circuit 242 is given as an input to a scaling amplifier 244, the gain of which is selected such that it generates an output signal which, after multiplication by the output value of the corresponding switch 234i to 234_N a code component at the output of the corresponding multiplier 236i to 236_N which ensures that the multiplied code component is masked by the selected part of the audio signal passed by the masking filter 240 when the encoded audio signal is reproduced as sound. Each weight factor determination circuit 230i to 230_N therefore generates a signal which represents an assessment of the suitability of the selected part of the audio signal for obscuring the corresponding code component.

In other embodiments of analog encoders according to the present invention, there are multiple weighting factor determination circuits for each code component generator, and each of the multiple weighting factor determination circuits corresponding to a given code component evaluates the suitability of a different part of the audio signal for masking that particular component when the encoded audio signal is reproduced as sound will. For example, there may be a number of such weighting factor determination circuits, each of which evaluates the suitability of a part of the audio signal for masking the respective code component when the encoded audio signal is reproduced as sound within a relatively narrow frequency band (so that the audio signal energy within such a band is large Probability consists of a single frequency component). A further weighting factor determination circuit can also be provided for the same respective code component in order to assess the suitability of the audio signal energy within a critical band, which has the code component frequency as the middle frequency, for concealing the code component when the encoded audio signal is reproduced as sound.

In addition, although the various elements of the embodiment of Figures 8 and 9 are implemented using analog circuitry, the same functions performed by such analog circuitry can also be implemented in whole or in part in digital circuitry.

Decode

In the following, decoders and methods for decoding will be discussed which are particularly suitable for decoding audio signals which are encoded by the above-described techniques according to the invention, as well as generally for decoding audio signals

AT 410 047 B so that the codes can be distinguished therefrom based on the amplitude. In accordance with certain features of the present invention and with reference to the functional block diagram of FIG. 10, the presence of one or more code components in an encoded audio signal is determined by establishing an expected amplitude. of amplitudes for the one or more code components, based on the audio signal level and / or a noise level that is not an audio signal, as indicated by the functional block 250. One or more signals representative of such expected amplitude or amplitudes are provided, as indicated at 252 in FIG. 10, to determine the presence of the code component by detecting a signal which corresponds to the expected amplitude or amplitudes. corresponds to the amplitudes as indicated by functional block 252. Decoders according to the present invention are particularly well suited for detecting the presence of code components which are obscured by other components of the audio signal, since the amplitude relationship between the code components and the remaining audio signal components is predetermined to some extent.

Figure 11 is a block diagram of a decoder in accordance with an embodiment of the present invention using digital signal processing to extract codes from encoded audio signals obtained by the decoder in analog form. The decoder according to Fig. 11 has an input connection 260 for receiving the coded analog audio signal, which can be, for example, a signal picked up by a microphone or a television or radio broadcast that is reproduced as sound by a receiver, or finally also coded analog audio signals that are in the form of electrical signals come directly from such a receiver. Such encoded analog audio signals can also be generated by reproducing a sound recorded in the form of a compact disc or tape cassette. Analog conditioning circuits 262 are connected to input 260 to receive the encoded analog audio signal and are used to perform signal amplification, automatic gain control and anti-aliasing low-pass filtering prior to analog-to-digital conversion. In addition, the analog conditioning circuits 262 are used to carry out a bandpass filtering process in order to ensure that the signals output thereby are limited to a frequency range in which the code components can occur. The analog conditioning circuits 262 deliver the processed analog audio signals to an analog-to-digital converter (A / D) 263, which converts the received signals into digital form and forwards them to a digital signal processor (DSP) 266 which processes the digitized analog signals to detect the presence of code components and the code symbols which they represent. The digital signal processor 266 is connected to a memory 270 (containing both program and data storage areas) and to input / output (I / O) circuits 272 to receive external commands (e.g. a command to initiate decoding or a command to output stored data Codes) and to output decoded messages.

The mode of operation of the digital decoder according to FIG. 11 for decoding audio signals encoded by means of the device according to FIG. 3 is described below. The analog conditioning circuits 262 are used for band filtering the coded audio signals, with a passband extending from approximately 1.5 kHz to 3.1 kHz, and the DSP 266 samples the filtered analog signals at a correspondingly high frequency. The digitized audio signal is then separated into frequency component domains (bins) by the DSP 266 using FFT processing. In particular, an overlapping, slice-like (window) FFT is applied to a predetermined number of least recent data points so that a new FFT is carried out periodically upon receipt of a sufficient number of new samples. The data is weighted, as will be explained below, and the FFT is carried out to generate a predetermined number of frequency ranges each having a predetermined width. The energy B (i) of each frequency range in a range including the code component frequencies is calculated by the DSP 266.

Noise level estimation is carried out around each area in which a code component may occur. If the decoder according to FIG. 11 is used to decode signals encoded by the embodiment according to FIG. 3, there are accordingly 40 frequency ranges within which a code component can occur. For everyone

AT 410 047 B such a frequency range, a noise level is estimated as follows. First, an average energy E (j) in the frequency ranges within a section or window which is in the frequency above and below the frequency range j of particular interest (i.e. the frequency range in which the code component can occur) is calculated according to the following relationship:

H (;) =

2w + 1

Xβ (/).

where i = (jw) -> (j + w) and w indicates the extent of the section above and below the area of ​​interest in numbers of areas. Then there becomes a noise level. NS (j) in frequency range j estimated using the following formula:

NS (j) = (ΣΒη (ί)) / (Σδ (ί)) where Bn (i) equals B (i) (the energy level in area i) if B (i) &lt;E (j) and B ( i) is otherwise equal to zero, and where δ (ϊ) is equal to 1 if B (i) &lt;E (j) and δ (i) is otherwise equal to zero.

That is, the noise components are believed to include those components whose level is below the average energy level within the particular section surrounding the area of ​​interest, and therefore include audio signal components which fall below this average energy level.

When the noise level for the region of interest is estimated, a signal-to-noise ratio SNR (j) for that region is estimated by dividing the energy level B (j) in the region of interest by the estimated noise level NSü). The values ​​of SNR (j) are used to detect both the presence and the timing (timing) of the synchronization symbols and the state of data symbols, as explained below. Various techniques can be used to eliminate audio signal components from consideration as code components on a statistical basis. For example, it can be assumed that the area with the highest signal-to-noise ratio includes an audio signal component. Another possibility is to exclude those areas with an SNR (j) value above a predetermined value. Another possibility is to eliminate from consideration those areas that have the highest and / or the lowest SNR (j) values.

When used to detect the presence of codes in audio signals encoded by the apparatus of FIG. 3, the apparatus of FIG a predetermined interval in which a code symbol can be found. The above procedure is accordingly repeated a number of times, and data relating to the presence of components is accumulated for each region of interest over this time frame.

Techniques for establishing appropriate detection time frames based on the use of synchronization codes are discussed in detail below. Once the DSP 266 has accumulated such data during the relevant time frame, it determines which of the possible code signals was present in the audio signal in the manner discussed below. The DSP 266 then stores the detected code symbol in the memory 270 concurrently with a time stamp for identifying the time the symbol was detected based on an internal timing signal of the DSP.

In response to a suitable command to the DSP 266 received via the I / O circuit 272, the DSP then causes the memory 270 to output the stored code symbols and time stamps via the I / O circuits 272.

The flowcharts of FIGS. 12A and 12B illustrate the sequence of operations that the DSP 266 performs in decoding a symbol encoded in the analog audio signal received at input terminal 260. Referring first to Fig. 12A, when the decoding process is initialized, the DSP 266 enters a main program loop at step 450, setting a SYNCH flag so that the DSP 266 first starts a process,

AT 410 047 B to detect the presence of the synchronization symbols E and S in the input audio signal in a predetermined message sequence. After performing step 450, the DSP 266 calls a DET subroutine, illustrated in the flow chart of Figure 12B, to look for the presence of code components representing the synchronization symbols in the audio signal.

Referring now to Figure 12B, in a step 454 the DSP repeatedly collects and stores samples of the input audio signal until a sufficient number is stored to perform the FFT described above. As soon as this has been carried out, the stored data are subjected to a weight function, for example a quadratic cosine weight function, a Kaiser-Bessel function, a Gaussian or Gaussian function. Poisson function, a Hanning function or some other suitable weight function, as indicated with step 456, in order to form a data segment. However, if the code components are sufficiently clear, weighting is not necessary. The segment-like data is then subjected to an overlapped FFT, as indicated in step 460.

Once the FFT is performed, a step 462 tests the SYNCH flag to see if it is set (in which case a sync symbol is expected) or reset (in which case a data bit symbol is expected). Since the DSP initially sets the SYNCH flag in order to detect the presence of code components which represent synchronization symbols, the program goes to a step 466 in which the frequency domain data obtained by means of the FFT in step 460 are evaluated to determine whether this data indicates the presence of components that represent an E-sync symbol or an S-sync symbol.

For the purpose of detecting the presence and timing of synchronization symbols, the sum of the values ​​of SNR (j) is first determined for each possible synchronization symbol and data symbol. At a given point in time during the process of detecting the synchronization symbols, a particular symbol is expected. The first step in detecting the expected symbol is to determine whether the sum of its corresponding SNR (j) values ​​is greater than any of the others. If so, a detection threshold is established based on the noise levels in the frequency ranges that may contain code components. That is, since only one code symbol is contained in the encoded audio signal at any given time, only a quarter of the regions of interest contain code components. The remaining three quarters contain noise, that is, program audio components and / or spurious energy. The detection threshold is generated as the mean value of the values ​​SNR (j) for all 40 frequency ranges of interest, but can be adjusted with a multiplication factor in order to take into account the effects of ambient noise and / or to compensate for an observed error rate.

With the detection threshold established in this way, the sum of the values ​​SNR (j) of the expected synchronization symbol is compared with the detection threshold to determine whether or not it is greater than this threshold. If so, a valid detection of the expected synchronization symbol is determined. Once this is done, as indicated in step 470, the program returns to the main process loop of FIG. 12A returns to step 472 where it is determined (as discussed below) whether a sample of the decoded data meets predetermined qualification criteria. If not, processing returns to step 450 to restart the search for the presence of a sync symbol in the audio signal; but if these criteria are met, it is determined whether the expected synchronization pattern (ie the expected sequence of symbols E and S) has been fully received and detected, as indicated in step 474.

However, after the first pass through the DET subroutine, insufficient data has been collected to determine whether the pattern meets the qualification criteria, so processing returns from step 474 to the DET subroutine for another FFT and evaluation for the presence of a synchronization symbol. After the DET subroutine has been run a predetermined number of times and when processing returns to step 472, the DSP determines whether the accumulated data meets the qualification criteria for a synchronization pattern.

This means that when DET has been passed through the predetermined number of times, one

AT 410 047 B corresponding number of evaluations have been carried out in step 466 of the DET subroutine. The number of times an E-symbol has been found is used in one embodiment as a measure of the amount of E-symbol energy during the corresponding period of time. However, other measures of the E-symbol energy can also be used instead, such as the entirety of the E-range SNRs that exceed the average range energy. After subroutine DET is called again and another rating is performed at step 466, at step 472 this least recent rating is added to those that were accumulated during the predetermined interval and the oldest among those that were previously accumulated will be deleted. This process continues for several passes through the DET subroutine and in step 472 a peak value in the E symbol energy is sought. If such a peak is not found, it will result in a determination that a synchronization pattern was not encountered, so processing returns from step 172 to step 450 to re-set the SYNCH flag and start the search for a synchronization pattern again.

If, however, such a maximum of the E-signal energy has been found, the evaluation process performed in step 472 after subroutine DET 452 continues, the same number of evaluations from step 466 being used each time, but the oldest evaluation being deleted and the most recent is added so that a sliding data window is used for this purpose. If this process continues, after a predetermined number of iterations in step 472 it is determined whether a crossover from the E symbol to the S has occurred. This is determined, in one embodiment, as the point at which the total of the SNR's of the S range resulting from step 466 within the sliding window first exceeds the total of the SNR's of the E range during the same interval. Once such a crossover point has been found, processing continues as described above to search for a maximum in S-symbol energy indicated by the greatest number of S-acquisitions within the sliding data window. If such a maximum is not found, or otherwise the maximum does not occur within an expected time frame after the maximum in E symbol energy, processing loops from step 472 back to step 450 to restart the search for a synchronization pattern.

If the aforementioned criteria are met, the presence of a synchronization pattern is determined in step 474 and processing moves to step 480 to determine the expected bit intervals based on the E and S symbol energy maxima and the detected crossover point. Instead of the method described above for detecting the presence of the synchronization pattern, other strategies can also be used. In a further embodiment, a synchronization pattern that does not meet criteria such as those described above, but that approximates a qualifying pattern (ie the detected pattern is not unambiguously eliminated), a determination of whether or not the synchronization pattern has been detected can be postponed, further analysis being based on evaluations (as described below) carried out to determine the presence of data bits at expected data intervals that follow the potential synchronization pattern. On the basis of the totality of the collected data, ie A retrospective qualification of the possible synchronization pattern can be carried out both during the dubious synchronization pattern interval and during the dubious bit interval.

Referring again to the flowchart of FIG. 12A, after qualifying the synchronization pattern in step 180, as mentioned above, the bit timing is determined on the basis of the two maxima and the crossover point. This means that these values ​​are averaged to determine the expected start and end points of each subsequent data bit interval. As soon as this has occurred, the SYNCH flag is reset in a step 483 to indicate that the DSP is then looking for the presence of one of the two possible bit states. The DET 452 subroutine is called again and, referring to Fig. 12B, the subroutine is executed in the same manner as described above up to step 462 in which the state of the SYNCH flag indicates that a bit condition should be determined,

AT 410 047 B and processing then goes to step 486. In step 486 the DSP looks for the presence of code components indicating either a bit state of 0 or a bit state of 1 in the manner described above.

Once this has been done, processing returns at step 470 to the main process loop of Figure 12A at step 490 where it is determined whether enough data has been obtained to determine the bit status. This requires several passes through the subroutine 452, so that after the first pass, processing returns to a subroutine DET 452 in order to carry out a further evaluation on the basis of a new FFT. As soon as the subroutine 452 has been run through a predetermined number of times, in step 486 the data collected in this way are evaluated in order to determine whether the data obtained indicate either a state of 0, a state of 1 or an indefinite state (which could be solved using parity data). This means that the total of the 0-range SNRs is compared with the total of the 1-range SNRs. The larger of the two values ​​determines the data state, and if the values ​​are the same, the data state is indeterminate. In an alternative case, if the total of the 0-range SNRs and 1-range SNRs are not the same but rather close together, an indefinite data state can be explained. Also, when a larger number of data symbols are used, the symbol for which the highest SNR summation has been found is determined as the obtained symbol.

When processing returns to step 490 again, the determination of the bit status is detected and processing proceeds to step 492 in which the DSP stores data in memory 270 indicating the status of each bit for composing a word that is a predetermined number of times of symbols which are represented by the coded components in the recorded audio signal. Thereafter, in a step 496, it is determined whether the recorded data has provided all bits of the coded word or of the message. If not, processing returns to the DET routine 452 to determine the bit state of the next expected message symbol. If, however, in step 496 it is determined that the last symbol of the message has been picked up, processing returns to step 450 to set the SYNCH flag and search for the presence of a new message by detecting the presence of its synchronization symbols as represented by the code components of the encoded audio signal.

Referring now to Figure 13, in certain embodiments, non-code audio signal components and / or other noise (collectively referred to as noise in this context) are used to generate a comparison value, such as a threshold, as provided by the functional block 276 stated. One or more parts of the encoded audio signal are compared with the comparison value, as indicated by the functional block 277, in order to detect the presence of code components. Preferably, the encoded audio signal is first processed to isolate components within the frequency band or bands that may contain code components, and these are then accumulated over a period of time to average out noise, as indicated by functional block 278.

Referring now to Figure 14, one embodiment of an analog decoder in accordance with the present invention is illustrated in block diagram form. The decoder of FIG. 14 includes an input terminal 280 which is connected to four groups of component detectors 282, 284, 286 and 288. Each group of component detectors 282 to 288 is used to detect the presence of code components in the input audio signal which represent a respective code symbol. In the embodiment according to Fig. 14th For example, the decoder is arranged to detect the presence of any of 4N code components, where N is an integer, so that the code consists of four different symbols each represented by a unique group of N code components. Accordingly, the four groups 282 through 288 contain 4N component detectors.

An embodiment of one of the 4N component detectors of groups 282 to 288 is shown in the form of a block diagram in FIG. 15 and is designated there as component detector 290. The component detector 290 has an input 292 which is connected to the input 280

AT 410 047 B of the decoder of Fig. 14 is connected to receive the encoded audio signal. The component detector 290 has an upper circuit branch with a noise estimation filter 294 which, in one embodiment, takes the form of a bandpass filter having a relatively wide passband for passing audio signal energy within a band centered on the frequency of the particular code component to be detected is. Alternatively and preferably, the noise estimation filter 294 instead comprises two filters, one of which has a passband extending from above the frequency of the particular code component to be detected, and a second filter having a passband with an upper margin below the frequency the code component to be detected, so that the two filters allow energy to pass through together, whose frequencies are above and below the frequency of the component to be detected, the frequencies not including this frequency, but being adjacent to it. An output of the noise estimation filter 294 is connected to an input of an absolute value circuit 296 which generates an output signal which represents the absolute value of the output signal of the noise estimation filter 294 and which is applied to the input of an integrator which accumulates the signals at its input to produce an output value representing the signal energy within parts of the frequency spectrum adjacent to, but not including the frequency of the component to be detected, and which outputs this value to a non-inverting input of a differential amplifier 202, which operates as a logarithmic amplifier.

The component detector according to Fig. 15th further comprises a lower branch which contains a signal estimation filter 306 which has an input which is connected to the input 292 for receiving the encoded audio signal and which serves to pass a band of frequencies which is substantially narrower than the relatively wide band of the noise estimation filter 294, so that the signal estimation filter 206 passes signal components essentially only at the frequency of the code signal component to be detected. The signal estimation filter 306 has an output which is connected to an input of a further absolute value circuit 308 which is used to provide a signal at its output which represents an absolute value of the signal passed through the signal estimation filter 306. The output of the absolute value circuit 308 is connected to an input of a further integrator 310. Integrator 310 accumulates the values ​​output from circuit 308 to produce an output signal representing the energy within a narrow pass band of the signal estimation filter over a predetermined period of time.

Both integrators 300 and 310 have a reset terminal which is connected to receive a common reset signal which is applied to a terminal 312. The reset signal is supplied from a control circuit 314 illustrated in Fig. 14, which regularly generates the reset signal.

Referring again to FIG. 15, the output of integrator 310 is applied to an inverting input of amplifier 302 which acts to produce an output signal equal to the difference between the output of integrator 310 and that of integrator 300. Since the amplifier 302 is a logarithmic amplifier, the range of possible output values ​​is compressed to reduce the dynamic range of the output in order to pass it to a window comparator 316 to detect the presence or absence of a code component during a given interval, which is determined by the control circuit 314 via the output of the reset signal. The window comparator is in the event that the input signal supplied by the amplifier 302 falls between a lower threshold, which is applied as a fixed value to an input terminal of the comparator 316 for the lower threshold, and a fixed upper threshold which is applied to an input terminal of the comparator 316 for the upper threshold is present, a code presence signal is generated.

Referring again to Figure 14, each of the N component detectors 290 from each component detector group provides the output of its respective window comparator 316 to an input of a code determination logic circuit 320. Circuit 320, under the control of control circuit 314, collects the various code presence signals from the 4N Component detector circuits 290 for multiple numbers of reset cycles as established by control circuit 314. Upon completion of the interval for the detection of a given symbol, which is established as described below, the

AT 410 047 B

Code determination logic circuit 320 determines which code symbol has been obtained as the symbol for which the greatest number of components have been detected during the interval, and outputs a signal indicative of the detected code symbol to an output terminal 322. The output signal can be stored in a memory, inserted into a larger message or file, sent or otherwise used (e.g. as a control signal).

The symbol detection intervals for the decoders described above in connection with Figures 11, 12A, 12B, 14 and 15 can be established based on the timing of synchronization symbols which are transmitted with each encoded message and which have a predetermined duration and sequence. For example, an encoded message contained in an audio signal can consist of two data intervals of the encoded E symbol followed by two data intervals of the encoded S symbol, both as described above in connection with FIG. The decoders of Figures 11, 12A, 12B, 14 and 15 initially operate to search for the presence of the first suspected synchronization symbol, ie according to the encoded E symbol transmitted during a predetermined period of time and its transmission interval. The decoders then search for the presence of the code components which characterize the symbol S and, as soon as this is detected, determine its transmission interval. The transition point from the E symbol to the S symbol is determined from the detected transmission intervals, and from this point the detection intervals are set for each of the data bit symbols. During each detection interval, the decoder collects the code components in order to determine the particular symbol which is transmitted during this interval in the manner described above.

Although various elements of the embodiment of Figures 14 and 15 are implemented with analog circuitry, it should be noted that the same functions performed in these circuits can be implemented in whole or in part using digital circuitry.

Referring now to Figures 16 and 17, there is illustrated a system for generating audience estimates from widely broadcast information, such as in the case of television and radio programs. Figure 16 is a block diagram of a broadcast station for wirelessly broadcasting audio signals encoded to identify the station along with a time of broadcast. If desired, the identity of a program or segment being broadcast can also be included. A program audio source 340, such as a compact disc player, a digital tape recorder or even a live sound source, is controlled by the station manager by means of a control device 342 in order to output the audio signals to be transmitted in a controlled manner. An output 344 of the program audio source is connected to an input of an encoder 348 in accordance with the embodiment of FIG. 3 which includes the DSP 104, the bandpass filter 120, the analog-to-digital converter (A / D) 124, the digital-to-analog converter ( DAC) 140 and summing circuit 142. The controller 342 includes the main processor 90, the keypad 96 and the monitor 100 of the embodiment of FIG. 3 so that the main processor residing within controller 342 is connected to the DSP contained in encoder 348 of FIG. The encoder 348 operates under the control of the controller 342 to periodically insert an encoded message into the audio signal to be transmitted, the message including appropriate identification data. The encoder 348 sends the encoded audio signal to the input of a radio transmitter 350, which modulates a carrier wave with the encoded program audio signal and transmits this wirelessly by means of an antenna 352. The main processor contained in the control device 342 is programmed via the keyboard to control the encoder so that it outputs the appropriate coded message including the identification data of the station. The main processor automatically generates the transmission time data by means of a reference timer circuit contained therein.

Referring now also to FIG. 17, a person monitor 380 of the system is enclosed in a housing 382 that is sufficiently small for an individual who is a member of an audience and an audience to be able to carry it participates. Each of a number of members of the audience has a personal monitoring device, such as device 380, which each person who is a member of the audience must wear at certain times of each day during the monitoring period, such as a predetermined time of one week . The person monitoring device 380 contains a microphone which is sensitive in all directions

AT 410 047 B

386 which picks up sounds heard by the audience member wearing device 380, including radio programs reproduced as sounds from the speaker of a radio receiver such as radio receiver 390 in FIG.

The person monitoring device 380 furthermore has a signal conditioning circuit 394, the input of which is connected to an output of the microphone 386 and is used to amplify the output of the microphone 386 and to subject it to a band filtering, both in order to attenuate frequencies that are outside an audio frequency band that the different frequency components of the audio signal in the program by the encoder 348 of Fig. 16 included codes, as well as to perform anti-aliasing filtering prior to analog / digital conversion.

The digital circuit of the person monitoring device 380 is shown in functional block diagram form in FIG. 16, with a decoding block and a control block, both of which can be implemented, for example, by means of a digital signal processor. A program and a data memory 404 is connected both to the decoder 400 in order to receive the detected codes for storage and to the control block 402 in order to control the writing and reading operations of the memory 404. An input and output circuit 406 (I / O) is connected to the memory 404 in order to receive data to be output from the personal monitoring device 380 and also to store information such as program instructions therein. The I / O circuit 406 is also connected to the control block 402 in order to control the input and output operations of the device 380.

The decoder 400 operates in accordance with the decoder of FIG. 11 described above and outputs station identification and time code data to be stored in memory 404. The personal monitoring device 380 furthermore has a connection, which is indicated schematically by 410, in order to output the collected station identification and time code data stored in the memory 404 and also to receive commands from an external device.

The personal monitoring device 380 is preferably capable of operation with the docking station disclosed in U.S. Patent Application Serial No. 08 / 101,558 (filed on

2. August 1993) entitled Compliance Incentives for Audience Monitoring / Recording Devices, the disclosure of which is incorporated herein by reference. In addition, the personal monitoring device 380 is preferably provided with the further features of the portable radio exposure monitoring device, which is also described in the cited US patent application.

The docking station communicates by means of a modem over telephone lines with a centralized data processing system in order to deliver the identification and time code data there in order to generate reports regarding the audience level of the audience. The centralized facility may also download information to the docking station for its own use and / or for the device 380, such as executable program information. The centralized system can also pass on information to the docking station and / or the device 380 via a radio frequency channel, for example an existing FM radio, which is encoded with this information according to the type of the present invention. The docking station and / or device 380 is provided with an FM receiver (not shown for the sake of simplicity and clarity) which demodulates the encoded FM broadcast for transmission to a decoder in accordance with the present invention. The encoded FM broadcast can also be provided via cable or other transmission medium.

In addition to monitoring by personal monitoring units, stationary units can be used, for example add-on units. The set-top units can be connected to receive the encoded audio signal in electrical form from a receiver, or they can use a microphone, such as microphone 386 of FIG. 17. The tower units can then monitor channels selected using the present invention, in addition to which they can optionally monitor the composition of the audience.

There are other applications for the coding and decoding techniques of the present invention. In one application, the soundtracks of commercials are provided with codes for identification in order to enable the advertising to be monitored

AT 410 047 B ensures that commercials have been broadcast by television, radio or in any other form at times agreed in advance.

In other applications, control signals are transmitted in the form of codes generated in accordance with the present invention. In one such application, an interactive toy receives an encoded control signal contained in the audio portion of a television, radio, or sound recording and decodes it, whereupon it takes action in response. In another application, parental control codes are included in the audio portions of television, radio or sound recordings so that, by decoding these codes, a receiving or reproducing device can perform a certain parental control or control function in order to selectively receive or reproduce Prevent broadcasts or recordings. Control codes can also be incorporated into cellular telephone communications to prevent unauthorized access to the use of cellular telephone IDs. In other applications, codes are inserted into telephone transmissions in order to distinguish between voice and data transmissions and to control the choice of a transmission path in a suitable manner in order to avoid the destruction of transmitted data.

Furthermore, various transmitter identification functions can be implemented, for example to ensure the authenticity of military transmissions or voice communication with aircraft. Applications for monitoring purposes are also possible. In one such application, research study participants wear personal monitoring devices that receive encoded messages that are added to public announcements or similar audio signals in retail stores or department stores to record the attendance of the participants. In other embodiments, employees wear personnel monitors that receive coded messages added to audio signals at the workplace to monitor the presence of personnel at particular locations.

Furthermore, secure communications can be carried out using the coding and decoding techniques of the present invention. In such an application, secure underwater communications are carried out by encoding and decoding according to the present invention either by assigning code component levels so that the codes are obscured by the surrounding underwater noise or by a noise source originating at the location of the code sender. In other applications, secure paging is accomplished by inserting masked codes into other wireless audio signal transmissions that are received and decoded by a pager.

The coding and decoding techniques of the present invention can also be used to authenticate voice signatures. For example, in an application for orders placed by telephone, a stored speech sample can be compared with an actual voice; in another example, data such as a security number and / or the time of day can be encoded and combined with a spoken utterance and then decoded, the data can be used for an automatically controlled processing of the spoken utterance. The coding device in this application can either be an additional device for a telephone or for some other voice communication device, or it can be a separate, fixed unit that is used when the spoken utterance is stored directly without it being transmitted via telephone lines or in is sent out in any other way. Another application is the provision of an authentication code in the memory of a portable telephone so that the spoken language contains the authentication code and thereby enables the detection of unauthorized transmissions.

It is also possible to make better use of the bandwidth of communication channels by inserting data into voice or other audio transmissions. In such an application, data indicating readings from aircraft instruments are inserted into the voice transmissions between the aircraft and the ground station to inform the ground control posts of the aircraft's operating status without the need for separate voice and data channels. The code levels are chosen so that the code components are covered by the speech transmissions so that interference is avoided.

AT 410 047 B (magnetic) tape piracy, i.e. the unauthorized copying of copyrighted works such as audio or video recordings and music, can also be detected by encoding a unique identification number into the audio portion of each authorized copy using the encoding technique of the present invention. If the encoded identification number is discovered on multiple copies, it is evident that unauthorized copying has taken place.

In another application, the programs are determined which have been recorded using a VCR including a decoder according to the present invention. Video programs, such as entertainment programs, commercials, etc., are encoded in accordance with the present invention with an identification code which identifies the program. When the VCR is put into the recording mode, the audio portions of the signals to be recorded are fed to the decoder to detect the identification codes contained therein. The detected codes are stored in a memory of the VCR for later use in preparing a record usage report.

Data indicative of copyrighted works transmitted by a broadcaster or otherwise by a service company may be collected using the present invention to ensure liability for copyright fees. The works are coded with a respective identification code in order to clearly identify them. A monitoring unit that monitors the broadcast or other broadcasts. transmitted signals are supplied via one or more stations or service providers, outputs the audio parts of the signals to a decoder according to the present invention, which detects the identification codes contained therein. The detected codes are stored in a memory in order to generate a report which can be used to determine the fees or license income.

Decoders that have been proposed according to the Motion Picture Experts Group (MPEG) 2 standard already incorporate some elements of the acoustic enhancement processing required to extract the data encoded according to the present invention, so that techniques for recording lock (e.g. prevent unauthorized recording of works protected by copyright) using codes according to the present invention, work well with MPEG 2 decoders. A suitable decoder according to the present invention is provided in the recording device or as an additional part thereto and detects the presence of a copy protection code in an audio signal fed to a recording. The recording device responds to the blocking code detected in this way by preventing the recording of the corresponding audio signal and any accompanying signal, for example a video signal. Copyright information encoded in accordance with the present invention is on-tape, does not require additional synchronization or timing, or readily accompanies program material.

In other applications, programs transmitted wirelessly, by cable, or in any other way, or in other programs recorded on tape, disk, or in any other way, contain audio that is encoded with control signals for use by devices controlled by one or more viewers or listeners are served. For example, a program that maps the path to be traveled by a cyclist includes an audio portion that is encoded in accordance with the present invention with control signals suitable for use with a stationary exercise bike to control the pedal resistance or the driving resistance according to the gradient of the mapped path to serve. When the user steps on the pedals on the stationary bicycle, he or she sees the program on a screen or other monitor, and the audio portion of the program is played back as sound. A microphone on the stationary bicycle transmits the reproduced sound, and a decoder according to the present invention detects the control signals contained therein and outputs them to a control unit for the pedal resistance of the exercise bicycle.

From the foregoing it will be seen that the techniques of the present invention can be implemented in whole or in part using analog or digital circuitry and that all or part of the signal processing functions can be implemented either by hardwired circuitry or by using digital signal processors, microprocessors, microcomputers , Multiple processors (e.g. parallel processors) or

AT 410 047 B similar can be carried out.

The features of the invention disclosed in the preceding description, in the drawing and in the claims can be essential both individually and in any combination for the implementation of the invention in its various embodiments.