Equalization of speech signal in a mobile phone

Speech signal frequency correction in a mobile phone - Talsignalens equalization in a mobile phone

The invention relates generally to the processing of a speech signal and in particular to frequency correction for a speech signal in a mobile telephone, the purpose of which is to improve the intelligibility of the transmitted speech.

As a special case for the voter, improving the hearing effect of speech with an equalizer or equalizer is a widely known method used.

For example, the frequency spectrum of sound may be equalized, with peaks or dips in the frequency response corresponding to certain frequencies being aligned closer to the average level, or certain frequencies or frequency ranges may be intentionally amplified or attenuated to deviate from the average level. In addition to improving the normal hearing effect, frequency correction can help people with certain types of hearing impairments to become clearer about their voice.

The frequency correction can be done electrically or mechanically. Electronic frequency correction requires the use of different filter circuits in the sound processing device. Known solutions in the field of analog signal processing are passive RC filters, active filters based on operational amplifiers and special equalizer chips, as well as digital signal processors on the digital side, which can be programmed to incorporate very complex filter arrangements. Mechanical frequency correction is performed by selecting the mechanical characteristics of the sound reproducing speaker and its immediate environment so as to achieve the desired frequency response.

In mobile phones, frequency correction is needed to improve the quality and intelligibility of the sound transmitted from the speaker to the user's ear. The sound from the speaker appears distorted because the speaker does not usually reproduce all frequencies in the same way and because there is an acoustic leakage between the speaker and the user's ear that affects different frequencies differently. Figure 1 shows graphically two frequency response measurements, of which in the measuring arrangement corresponding to the upper curve the loudspeaker and the corresponding measuring device are protected against acoustic leakage (strong acoustic connection) and in the arrangement corresponding to the lower curve the acoustic leakage corresponds to normal mobile phone use. It can be clearly seen from the figure that when the acoustic coupling deteriorates due to leakage, the repetition of low frequencies (<1 kHz) in particular deteriorates.

One known method of trying to correct the sound distortion caused by acoustic leakage is to use special speakers that apply acoustic feedback. The method is mechanical in nature because it seeks to construct the speaker so that its mechanical properties compensate for low-frequency attenuation. An air gap is left between the speaker and its mounting frame, the width of which is an important factor for acoustic feedback. To achieve proper sizing and operation, precise mechanical tolerances must be adhered to in the speaker assembly, which increases manufacturing costs. The frequency response of a mechanical arrangement cannot be changed afterwards, which makes it inflexible. Nor can the method be applied to the implementation of a hands-free speaker, which is a common mobile phone accessory.

By designing the electromechanical properties of the speaker, it is also possible to implement a speaker that inherently amplifies low frequency signals. Such an arrangement is also inflexible, because the frequency response is determined during the manufacture of the loudspeaker and cannot be changed, for example, by an electronic control signal.

Electronic filter solutions can generally have a more versatile and flexible effect on the frequency response of a sound reproduction arrangement than mechanical or electromechanical20 solutions. In a mobile phone, however, space, power consumption, computing capacity, and circuit board area are very critical factors that an electronic filter should consume as little as possible. Both passive and active analog filters increase component costs and take up both electrical power and circuit board surface area. If they are to be used to build an adjustable equalizer, the operation of which can be changed with a control signal if necessary, the arrangement can easily become complicated and expensive to manufacture. Analog filters often also have a certain, rather limited operating range, such as a fixed center frequency or insufficient gain.

Digital filter arrangements, which can in principle be implemented as separate digital circuits, but whose preferred embodiment is to program them into processes to be performed by a digital signal processor, have proven to be superior in many respects compared to their analog counterparts. Since most common cellular telephone designs today already include a digital signal processor35, or DSP, a software filter arrangement does not significantly increase the number of cellular components. At the same time, electricity consumption and space requirements remain within reasonable limits. The operation of the processes programmed into the DSP can also be easily changed as desired by giving the processor certain control signals.

However, prior art DSP-based equalizer solutions for mobile phones do not work optimally in all situations. Namely, it is known that since the mobile phone is used in very different conditions, for example, the level of ambient noise in both speech and listening environments varies, it is not possible to reproduce speech in the best possible way if the frequency correction is always done in the same way. In addition to the environmental conditions, the use of various accessories, such as a hands-free speaker, a headphone speaker, or an additional headset, will affect the operation required of the equalizer.

It is an object of the present invention to provide a method and apparatus by which the quality and intelligibility of the sound reproduced by a mobile telephone can be improved in a variety of situations. It is also an object of the invention to provide a method and a device with which the quality and intelligibility of the sound reproduced by a mobile telephone can be controlled in connection with the use of various accessories. It is a further object of the invention to provide a method and apparatus for improving the sound reproduction of a mobile telephone which do not require the use of complicated and expensive special speakers. It is a further object of the invention that the method and apparatus according to it do not significantly increase the use of critical resources of a mobile telephone.

The objects of the invention are achieved by a solution in which the signal received by the mobile telephone is subjected to adaptive frequency correction after speech decoding but before D / A conversion of the signal. Adaptability means that the effect of the frequency correction is changed when certain control parameters change. The control parameters25 meters and their effect are described in more detail below.

The method according to the invention for improving the sound reproduction of a telephone device comprising an equalizer having a certain variable frequency response is characterized in that it generates control information describing the operating conditions of said telephone device and converts the frequency response of said frequency correction to the reproducible sound quality.

The invention also relates to a telephone in which the method according to the invention is applied. The telephone device according to the invention, comprising a loudspeaker for sound reproduction and digital signal processing means for digital processing of reproducible sound, whi ch digital signal processing means comprise an equalizer having a certain frequency response, that said telephone device kä99062 provides means for generating control information describing its operating conditions and means for changing the frequency response of said equalizer based on said control information.

The present invention is intended for use in a digital telephone, preferably a mobile telephone. The applicability of the invention requires that the telephone process the sound as a digital signal. The audio information received in a conventional mobile phone is in digital form before it is converted to analog form (PCM decoding; Pulse Coded Modulation) and for audio Reproduction on a loudspeaker. In addition, the audio information to be transmitted is in digital format when it is first recorded with a microphone and A / D converted (PCM encoding). The operation of a digital mobile phone per se is considered to be known to a person skilled in the art and will n ot be described in more detail here.

In the research work leading to the invention, it was found that the mere amplification of low frequencies is not recommended in all conditions for the intelligibility of the received speech. If the mobile phone is used in a very noisy environment, amplifying the low frequencies may even impair speech intelligibility, as it reduces the relative share of the higher, most intelligible frequency components of the signal heard by the ear. Thus, it is advantageous to relate the frequency correction to the ambient noise level. In addition, it has been found that because different speakers are used in mobile phone accessories, the frequency correction must be adjusted to work differently with them.

According to the invention, several different control parameters can be used to adjust the frequency correction, the simplest of which is the volume of the telephone set by the user. Other possible control parameters are e.g. the measu rement of the ambient noise level through the microphone of the telephone, information on the accessories connected to the telephone and the measurement results describing the quality of the radio connection. The final sound emitted by the user consists of an interaction consisting of the original signal, frequency correction, possible D / A bandpass filtering, gain, speaker acoustic response, and in normal telephone operation, acoustic leakage between speaker and ear or hands-free when using the hands-free function. .

In the method and device according to the invention, the mobile phone examines and determines, on the basis of the control parameters and the information stored in the memory, which is the best frequency correction force suitable for the respective environmental and operating conditions. Based on the determinations, the received signal is preferably frequency corrected in the same digital signal processor used in a continent known per se to perform speech de coding.

Advantages of the invention over the prior art include improved sound intelligibility regardless of speaker design and acoustic leakage between speaker and ear, good suitability for special requirements of mobile phones, adaptation of frequency correction to background noise and user-set phone volume, balance between audible sound quality and speech intelligibility.

The invention will now be described in more detail with reference to exemplary preferred embodiments and the accompanying drawings, in which frequency response measurements are shown graphically in the presence of strong and weaker acoustic coupling between a loudspeaker and a measuring device, a block diagram showing the principle of the invention and its application in a mobile phone. shows as a block diagram a filter that can be used in the equalizer according to the invention, graphically shows the frequency response scaling in the arrangement according to the invention, graphically shows certain control curves that can be applied in the arrangement according to the invention, and shows as a block diagram the estimation of ambient noise for use according to the invention.

In the above description of the prior art, reference has been made to Figure 1, so in the following description of the invention and its preferred embodiments, reference will be made mainly to Figures 2-9. In the figures, the same reference numerals are used for the corresponding parts.

Figure 2 shows a block diagram, the two main blocks of which are a digital signal processor (DSP) 1 responsible for digital processing of a speech signal in a mobile telephone and a PCM CODEC block 2 responsible for PCM coding and decoding of a signal and analog signal processing. far-end channel, i.e. the signal path from the mobile phone receiver and channel decoding (not shown in the figure) through the speech decoder 3, the equalizer block 4, the D / A converter 5, the filter 6 and th e amplifier / attenuator 7 to the loudspeaker 8. a near-end channel, ie, a signal path from the microphone 9 through an amplifier 10, a filter 11, an A / D Converter 12, a background noise measuring block 13, and a speech Encoder 14 toward channel coding (not shown) and transmission to the radio path. With the exception of the equalizer 4 and background noise measuring blocks 13 according to the invention, the construction and operation of all the listed parts are within the competence of a person skilled in the art, since the corresponding parts are commonly used in prior art mobile telephones. The blocks according to the invention also consist of parts known per se to a person skilled in the art, but their connection to a mat20 intercom and their use are within the scope of the present invention.

The DSP block 1 comprises a control block 15, the main function of which for the present invention is to control the equalizer 4 so that signal components of a certain freque ncy are amplified or attenuated according to the prevailing environmental and operating conditions. The operation of the control block 15 is based on the control parameters that enter the control block 15 as input data. All control parameters in some way describe the environmental and operating conditions in which the mobile phone is used. Figure 2 shows four control parameters as input data for the control block 15, which are information about the quality of the radio connection 16, background noise based on ambient noise measurement 17, information 18 about accessories connected to the telephone and user-set volume 19. The figure also shows memory means or look-up tables. 4a and 13e, the use of which is described below.

The frequency response of the equalizer 4 can vary in many ways. Figure 3 shows by way of example a variable gain when the center frequency and the bandwidth are constant. Accordingly, Fig. 4 shows an exemplary variable center frequency when the gain and bandwidth are constant, and Fig. 5 shows an exemplary variable bandwidth when the center frequency and gain are constant. Changes in center frequency, gain, and bandwidth can be combined in many ways depending on how to get the best result, i.e. the clearest and most understandable sound. The operations preferably take place in the normal speech frequency range of about 300 to 3400 Hz. A special case of the invention can be considered a constant frequency correction, which is not adaptive but always works in the same way.

The equalizer block 4 may consist, for example, of one or more IIR filters (Infinite Impulse Response filter, recursive filter), FIR (Finite Impulse Response filter, non-recursive filter) or a process operating in frequency cut-off. In the latter case, block 4 must also contain time-frequency and frequency-time transformations, which are most naturally implemented with FFT and IFFT (Fast Fourier Transform, Inverse Fast Fourier Transform) functions. <

Figure 6 shows a preferred implementation of the equalizer block. In this filter structure known per se, the blocks marked with Z denote delays of a unit of time, α1 and α2 are factors influencing the center frequency and bandwidth of the frequency response, and the K factor corresponds to the gain of the frequency correction. The presented filter is exemplary, because the detailed implementation of the equalizer block 4 is not essential per se for the invention. The advantages of the filter solution in Figure 6 are only three adjustable factors (ai, a2 and K), low effect on the signal phase, good and clear adjustability and good stability. The transfer function of the filter is of the formula

- (ai +1 + K (a \ -1)) + (a2 · Z '¹) + - (l + ai + (l - al)) z² + (α2 · Ζ ~ ') + (αΙ · Ζ-² adequate.

Next, the operation of the method and device according to the invention will be described. The basic idea of ​​Pe30 is th at the control block 15 is provided with information in the form of control parameters about the conditions and equipment in which the telephone is used, whereby the control block 15 controls the equalizer block 4 so as to achieve the best possible signal quality intelligibility.

The simplest control parameter is information about the user-set volume of the phone. Most modern cell phones have a controller with which the user

<img file="FI99062C_D0001.tif" />

may increase or decrease the volume of the loudspeaker. In normal use, the user adjusts the volume louder if there is a lot of noise in the operating environment, and quieter if the operating environment is quiet. As stated above, it is advantageous from the point of view of the quality of the received sound to emphasize low (about &lt;1 kHz) frequencies in a quiet operating environment, but from the point of view of speech intelligibility it is advantageous to attenuate them in a noisy operating environment. The control block 15 is provided with information on the position of the volume control in a manner known per se, whereby it changes the control information of the equalizer block 4. If the equalizer block according to Fig. 6 is used, the set volume 10 causes the control block 15 to decrease the factor K, and the correspondingly set low volume causes the control block to increase the same factor. In the preferred embodiment, the maximum possible value of the factor K corresponds to the largest integer in the read representation of the digital signal processor 1, whereby the gain cannot be increased so much as to cause a numerical overflow in the processor. Figure 7 schematically shows the gain of the equalizer block 4 as a function of frequency when the maximum possible value of the K-factor is used.

A more accurate picture of the background noise of the operating environment is obtained by measuring it through the microphone 9. Many modern mobile phones have so-called VAD (Voice Ac20 tivity Detection) function, which is used to detect during a call when the user is talking on his phone and when he is silent. Normally, this feature is used so that when the user is silent, the phone sends only a minimum amount of monitoring information, which reduces the power consumption of the phone and the load on the radio frequencies. The VAD function can be applied to the method and device according to the invention, because when the user is silent, the microphone 9 only records background noise. By measuring the background noise level at times when the VAD unit (not shown) notifies the control block 15 that the user is not speaking, and storing the measurement result for use as an estimate of the background noise level until the next measurement result is obtained, the equalizer block 4 can be controlled as described above. . This requires that the telephone, preferably its digital signal processor, comprise a background noise measurement block 13 from which the control block 15 receives the measurement results. In addition, the VAD unit (not shown in the figures) must be connected to the control block 15 in order to know when the measurement results can be received. The information provided by the VAD unit can also be fed to the background noise measuring block 13 in the nature of on / off control, whereby it provides measurement results to the control block 15 only when the user is silent according to the VAD unit.

In measuring the background noise, the power level of the signal received from the A / D converter 12 of the PCM block 2 is preferably monitored. It is advantageous to use some kind of averaging in the measurement of the power level, because sudden changes in the power level, especially rapid power peaks, could otherwise cause fluctuations in the frequency correction so fast that they would interfere with the hearing effect. The measurement technique uses the terms on and off time (attack and release time), which describe how quickly the system reacts to the rise of a signal to be measured above a certain threshold (attack) and to fall below it (release). In measuring the power level, it is advantageous to use a longer attack time to avoid interference caused by fast power peaks and a short release time so that the system reacts quickly enough to a sudden silence of background noise when, for example, the car in the operating environment is stopped.

From the user's point of view, the best way would be to measure the so-called background noise instead of the power level. loudness, which in large quantities describes the human moon15's impression of power level changes. However, in order to measure the loudness, complex filter arrangements and frequency-dependent calculations are required, whereby the performance of the digital signal processor 1 may become a limiting factor. However, new, more powerful processors may be sufficiently efficient in the future, which could be known as A-weighting, F-weighting filters.

The background noise measuring block 13 may also comprise frequency analysis means which measure not only the intensity of the background noise but also its frequency spectrum. Based on these measurement data, the frequency response of the equalizer block 4 can be changed so that the correction made to the speech signal specifically compensates for the effect of a certain frequency noise. This embodiment is particularly advantageous if there is a machine in the operating environment which, when running, produces noise weighted to a certain frequency.

Mobile telephones also generally have functions that measure the quality of the radio connection and on the basis of which the transfer of the connection is performed, i.e. the so-called handover from one base station to another. The results of these measurements can also be used in the method and device according to the invention for controlling the operation of the equalizer. The main principle in the preferred embodiment is that the frequency correction is intensified when the quality of the connection is found to deteriorate, for example in the form of a decreasing field strength or in the form of successive failed frames. Other types of remediation strategies are also possible, and the best course of action can be found by experimentation. The use of a radio link quality control parameter or parameters requires that a portion of the mobile telephone measuring the radio link quality be connected to the control block 15. This is easy to provide because in known implementations many link quality parameters are computed in the same digital signal processor affordable to invest.

Figure 2 shows the control parameter input numbered 18, which describes the accessories connected to the telephone. Certain accessories, such as headphones and handsfree speakers, require different frequency adjustments to achieve the best possible sound quality. In addition, when using a hands-free speaker, for example, the effect of the mobile phone volume control is stronger than in normal mobile phone use, which must be taken into account when the frequency correction control is otherwise based on the volume setting only, but the control block 15 also receives information about the connected accessory. In the preferred embodiment, the optimal values ​​of the control information of the equalizer block 4 corresponding to each accessory are measured by the telephone manufacturer and stored in the memory means used by the control block 15 or directly in the equalizer block 4 during programming. look-up table 4a. When the control block 15 receives as a control parameter information that a certain accessory is connected to the telephone, it reads the values ​​of the control data corresponding from the memory means and transmits them to the equalizer block 4.

In the method and device according to the invention, in addition to the above, other control parameters or similar means can be applied to change the operation of the equalizer block 4. For example, if the user has a hearing impairment that requires a certain type of frequency correction, the equalizer block 4 can be adjusted to make the correct type of correction. In a preferred embodiment of the invention, in which the functions shown inside the DSP block 1 are implemented as software processes, their operation can be easily influenced even on a user-by-user basis by programming the digital signal processor in the correct way.

The co-operation of the above control parameters must be adapted so that it is logical and suitable for varying operating situations and does not exceed the performance of the digital signal processor 1. In a preferred embodiment, the volume of the telephone set by the user has a direct relationship to the K-factor of the equalizer, i.e. the gain, so that even a small increase in volume decreases the Factor and vice versa. In Fig. 8, each of the curves A, B, and C schematically depicts the change in gain when the volume control is changed, but the other conditions remain constant. The background noise is measured in block 13 in steps, i.e. the measured background noise power level is compared with certain threshold values. Exceeding the threshold causes the K-factor to be scaled by a constant factor, which in Figure 8 corresponds to jumping from curve A, B, or C to another curve. For example, when the background noise increases above a certain threshold, it is possible to jump from curve B to curve C and, respectively, when the background noise subsides from curve A back to curve B or even directly to curve C.

The attack and release time factors associated with jumps are described above.

If the quality of the radio connection is so poor that the parameter describing it exceeds a certain threshold level, a certain standard gain can be introduced, which does not depend on the volume set by the user or the measured noise level. This is illustrated by the straight line D in Figure 8. As shown above, the poor connection quality can also be compensated by another type of correction strategy, in which case the best operating model can be found by experimentation. In addition to the cooperation modes shown, the look-up table 4a stores control information of the equalizer block 4 related to certain accessories, which bypasses the information intended for normal telephone use when a certain accessory is switched on. In this case, the parameters controlling the frequency correction can change considerably at once. During a call, when the equalization is controlled on the basis of background noise and volume control, sudden changes in the operation of the equalizer block 4 are most preferably avoided, as they could cause perceptible and disturbing changes in the sound heard by the user.

Figure 9 schematically shows a more detailed implementation of the background noise measuring block 13. The actual power level measurement takes place in block 13a, which may be preceded by a weighting filter 13b, especially if it is desired to weight the noise measurement instead of the power level in the direction of loudness. In block 13c, based on the sig25 signal provided by the VAD unit, it is examined whether the user is speaking or not, and the measured noise value is compared with the threshold value obtained from the look-up table 13d. An absolute value or a code value that refers only to certain threshold values ​​can be given as an estimate of the noise level.

The control methods shown, which are mainly based on threshold value comparisons, require that the digital signal processor 1 has a sufficient amount of memory, preferably ROM memory, in which threshold values ​​and other look-up table data can be stored. The number of different combinations and alternative control strategies is usually limited by the amount of ROM available. The control methods could be implemented even more flexibly if the control data of the equalizer block 4 were calculated in real time35, taking into account all the changing factors continuously. However, this requires a larger computational capacity, in which case the capacity and instruction set of existing digital signal processors based on integer arithmetic may become limiting factors. In the future, more powerful processors are likely to provide new opportunities in this regard.

A digital mobile telephone has been discussed above to illustrate the applicability of the invention. However, the invention is not limited to mobile telephones, but can be used in all telephone devices in which the reproducible sound is processed in digital form and in which some of the shown or equivalent control parameters can be used to describe the conditions to which the reproducible sound frequency correction must be adapted. In addition to mobile telephones, the invention is applicable, for example, to digital cordless telephones which have sufficient signal processing means to implement the functions required for frequency correction and its control. The control parameters presented herein are exemplary in the sense that any of them or any combination thereof may be used to implement the adaptive frequency correction of the invention. One skilled in the art can also present other functionally equivalent control parameters.

In a preferred embodiment of the mobile telephone according to the invention, in connection with the design of the PCM block 2, it is taken into account that the frequency-dependent operations performed in the equalizer block 4 affect the loudness perceived by the user.

In particular, the amplifier / attenuator unit 7 belonging to the PCM block must be designed in such a way that it can prevent any noise caused by frequency correction.^: unwanted change.

The method and device according to the invention meet the objectives set for them and represent a clear improvement over the prior art, since the adaptive frequency correction significantly improves both the quality and intelligibility of the audible sound according to the current use situation.