Efficient filtering with a complex modulated filterbank

(74) Representative (54) Designation Effective filtration with single complex, modulated filter bank (56) Published publications US 6236731 B1, US 5848108 A (57) Abstract

A filter apparatus for filtering a time domain's input signal to obtain a time domain's output signal which is a representation of a time domain's input signal filtered using a filter characteristic with a non-uniform amplitude / frequency characteristic comprises a complex analysis filter bank for generating several complex subband signals , multiple middle filters, wherein at least one of the intermediate filters of the multiple intermediate filters has a non-uniform amplitude / frequency characteristic, wherein the multiple intermediate filters have a shorter pulse response compared to a pulse response of a filter with the filter property and where the non-uniform amplitude / frequency property of the multiple intermediate filters together represents the non-uniform filter characteristic and a complex synthesis filter bank to synthesize the signal from the intermediate filters to obtain the output of the time domain.

<img file="NO342467B1_D0001.tif" />

Description

Discipline

The invention relates to a filter apparatus and method for filtering an input signal in the time domain, a filter converter and a method for generating an intermediate filter definition signal, especially in the area of ​​coding, decoding, manipulation and filtering of audio signals, i.e. in the area of ​​HRTF (head-related transmission function). ).

Background

It has been shown in [P. Ekstrand, Bandwidth extension of audio signals by spectral band replication, Proc. 1^st IEEE Benelux Workshop on Model Based Processing and Coding of Audio (MPCA-2002), pp. 53-58, Leuven, Belgium, 2002], that a complex exponential modeled filter bank is an excellent tool for spectral envelope tuning of audio signals. One application of this feature is audio coding based on spectral band replication (SBR). Other useful applications of a complex filter bank include frequency selective panning and spatialization for parametric stereo, see [E. Schuijers, J. Breebart, H. Purnhagen, J. Engdegård: Low complexity parametric stereo coding, Proc. 116^th AES convention, 2004, paper 6073] and multi-channel parametric coding, see [J. Herre et al .: The reference model architecture for MPEG spatial audio coding, Proc. 118^th AES convention, 2005, paper 6447], In these applications, frequency resolution of the complex filter bank is further enhanced at low frequencies by means of subband band filtering. The combined hybrid filte r bank thereby achieves frequency resolution which enables the processing of spatial instructions at a spectral resolution that closely follows the spectral resolution of the two channel audio system.

Furthermore, US6236731 B1 describes a filter bank structure that provides a flexible compromise between the conflicting goals of processing time delay, filter sharpness, memory usage and tape interaction. The filter bank has an adjustable number of bands and a stack that provides a selectable change of band frequencies to one of two discrete sets of center frequencies. The width of the tapes and thus the number of tapes is selected depending on the acceptable delay, memory usage and processing speed required. The flexibility of stacking the bands provides twice as many potential band edge positions, which is advantageous for hearing loss adaptation, especially at low frequencies. The same filter coefficients can be used for analysis and synthesis to reduce memory usage.

US 5848108 describes how complex signal samples of an input signal are supplied via a frequency downshift to a digital filter and to a time frequency converter unit which produces multiple frequency components each connected to a respective frequency band within a bandwidth of the input signal. The frequency component having the maximum instantaneous energy is determined to determine the down-shift frequency of the associated frequency band to be filtered, and to determine a complementary offset by the digital filter to produce an output signal. The filter may be a bandpass filter or a bandstop filter, specifically to reduce inter-channel interference in different types of wireless communication systems.

However, in some applications, the solution in the filter bank is still insufficient in that simple amplification modifications in each subband are not sufficient to properly model the effect of a given filter. For dual-channel reproduction of multichannel audio using HRTF (head-related transfer function) associated with filtering, the intricate phase properties of the filters are important for the perceived sound quality. Of course, it is possible to use fixed inference methods based on DFT (discrete Fourier transform) as post-processing in the multi-channel rendering, but if the rendering device already contains signals in the subband domain of a complex exponentially modeled filter bank, there will be significant advantages with computational complexity and algorithmic integration. the performance of HRTF-derived filtering in the subband domain as outlined in detail later. Since the HRTF is different for each individual and the derived filters depend on the virtual source and / or listener position such as. can be changed by control signals, user interfaces or by other description signals, it is also important to be able to effectively convert a given HRTF-related filter to subband domain filters.

Accordingly, it is an object of the invention to provide a filter apparatus for filtering an input signal in the time domain, a method for filtering an input signal in the time domain, a filter converter, or a method for providing an intermediate filter definition signal enabling a more efficient or more flexible manipulation an input signal in the time domain that provides better quality.

This object is achieved by a filter apparatus according to claim 1, by a method for filtering an input signal in the time domain according to claim 26, a filter system according to claim 25 or by a computer program according to claim 27.

Summary of the Invention

An embodiment of the invention relates to a filter apparatus for filtering an input signal in the time domain to obtain an output signal in the time domain which is a representation of the time domain input signal filtered using a filter characteristic with a non-uniform amplitude / frequency characteristic comprising a complex analysis filter. generate more complex subband signals from the time domain's input signal, multiple intermediate filters, for filtering the multiple complex subband signals to obtain a plurality of filtered complex subband signals, wherein at least one of the intermediate filters of the multiple intermediate filters has an intermediate non-uniform amplitude / frequency characteristic. Each intermediate filter of the multiple intermediate filters has a shorter pulse response compared to a pulse response with a filter having a non-uniform amplitude / frequency filter characteristic. The filter apparatus further comprises intermediate non-uniform amplitude / frequency characteristics of the multiple intermediate filters which together represent the non-uniform amplitude / frequency filter characteristic, and a complex synthesis filter bank for synthesizing the output of the intermediate filters to the multiple intermediate filters to obtain the output signal in the time domain.

In another aspect, a filter system is described for filtering the time domain's input signal to obtain the time domain's output signal comprising a filter apparatus according to any one of claims 1 to 24, to which the time domain's input signal is assigned as the time domain's input signal, and from which the time domain's output signal is obtained. filter system. The system further comprises a filter converter for providing an intermediate filter definition signal, comprising a complex modulated filter bank for filtering an impulse response signal indicating an amplitude / frequency filter characteristic in the time domain to obtain multiple subband signal with complex values ​​as the intermediate filter signal, with each subband signal being associated with each subband signal. modulated filter bank corresponds to an impulse response for a subband signal. At least one of the complex values ​​subband signals comprises at least two different non-vanishing values ​​and each complex value subband signal is shorter than the impulse response signal, the filter converter being coupled to the filter apparatus to provide multiple intermediate filters with an intermediate filter definition. The multiple intermediate filters of the filter apparatus are arranged to have impulse responses based on the intermediate filter definition signal.

In another aspect, a method for filtering the time domain input signal is described to obtain a time domain output signal which is a representation of a time domain input signal filtered using a filter characteristic having a non-uniform amplitude / frequency characteristic, comprising the steps of: Generating several complex subband signals complex filtering of the time domain's input signal; filtering each of the complex subband signals of the multiple complex subband signals using an intermediate filter of the multiple intermediate filters, to obtain more filtered complex subband signals, wherein at least one of the complex subband signals is filtered using an intermediate filter of the multiple intermediate filters with an intermediate non-uniform amplitude frequency characteristic. Each intermediate filter of the plurality of intermediate filters has an impulse response shorter in samples than the impulse response of a filter with the non-uniform amplitude / frequency characteristic of the filter characteristic; until the impulse response used for total filtering of the multiple subband signals represents the non-uniform filter characteristic. The method further comprises synthesizing the plurality of filtered complex subband signals emitted by the plurality of intermediate filters using a complex synthesis filter bank to obtain the time domain output signal.

Further, it is disclosed that the embodiments of the invention are based on the finding that a more efficient and / or more flexible filtering (or manipulation) of a time domain input signal can be obtained in the subband domain, and which is sometimes also called a quadrature mirror filter domain with a better quality compared to other manipulation systems. The gain in efficiency, especially the computational efficiency, is a consequence of the shorter pulse responses in the intermediate filters compared to the pulse response in a filter with a non-uniform filter characteristic in the time domain and that the subband signals can be processed independently of one another. Due to the shorter pulse responses, an embodiment of a filter apparatus can process the output of each complex subband signal individually by the complex analysis filter bank. Accordingly, the filtering can be performed in parallel, which dramatically increases the processing of the time domain's input signal as compared to the manipulation of the time domain's input signal directly, due to the shorter pulse responses.

Embodiments of the invention are particularly favorable when it comes to balancing computational efficiency on the one hand and quality on the other. While direct processing of the time domain's input signal in the time domain can be obtained by wrapping with the pulse response of a filter of the non-uniform amplitude / frequency characteristic which usually results in very good quality, the wrapping requires a large calculation of efficiency due to the length of the pulse response of the filter in the time domain. On the other hand, the conversion of an audio signal to the frequency domain by performing a Fourier transformation represents a major disadvantage in that other manipulations needed in modern acoustic systems cannot be effectively performed in the high-quality Fourier domain.

Using multiple intermediate filters, each having a shorter pulse response compared to a pulse response in a filter with the filter characteristic of a corresponding filter in the time domain where at least one has a pulse response with at least two non-vanishing values ​​representing a very favorable compromise between the computational efficiency of one page and the quality of the other side. As a result, the embodiments with the new filter apparatus represent an excellent compromise between a direct processing of the time domain's input signal e.g. using the envelope of the time domain's input signal with the longer pulse response indicating the non-uniform filter characteristic leading to a huge computational effort and using a Fourier transform which causes more problems in the further processing of the signals.

The advantages of the embodiments of the first aspect of the invention are particularly apparent in the context of the final impulse response (FIR) filters, as each of the intermediate filters of the multiple intermediate filters has a significantly shorter pulse response compared to the pulse response of the FIR filter in the time domain. Consequently, by parallel processing of the various subband signals of the complex analysis filter bank, the computational efficiency can be drastically improved. This aspect is especially important in the field of long pulse response filters. One area where filters with very long pulse response often occur is HRTF (head related transfer function), such as by mixing multi-channel audio signals for headphones, other head-related speaker systems or stereo sound systems.

In many specific applications, computational efficiency becomes even more efficient as audio signals are already present in the (complex) subband or QMF domain. In many specific implementations, the complex analysis filter bank and complex synthesis filter bank for generating several complex subband signals from the time domain's input signal and to synthesize the time domain's output signal are already present.

In the second aspect, the embodiments of the invention are based on the finding that a more flexible and more efficient filtering of the better-quality input signal of the time domain can be obtained by providing an intermediate filter definition signal such as e.g. may be provided to a filter apparatus according to the first aspect to define its intermediate filters.

A significant advantage of the embodiment of the second aspect of the invention is that an intermediate filter definition signal for a set of intermediate filters is obtained by providing an embodiment of the new filter converter with a filter defining signal, e.g. a pulse response signal indicating an amplitude / frequency filter characteristic of a filter in the time domain or other filter definition signals. Accordingly, an embodiment of a filter converter provides a filter definition signal for a set of intermediate filters to perform the same filtering as a filter in the time domain defined by the filter definition signal virtually without introducing side effects. As a result, the designs of the new filter converter make it possible to achieve a virtual alias-free performance of any filter in the subband domain. By utilizing an embodiment of the new filter converter, arbitrary filter properties can be transferred from the time domain to the subband signal domain, e.g. virtual alias-free smoothing, low-pass filter characteristics, high-pass filter characteristics, band-pass filter characteristics, tape rejection filter characteristics, resonance filter characteristics, suction filter characteristics, or more complex filter characteristics. Among the more complex filter characteristics, a combination of several characteristics as well as HRTF-related filter characteristics can be mentioned.

Particularly in the context of HRTF-related applications in the field of multi-channel audio systems and other high-quality applications, it is important to note that the designs of the new filter converter make it possible to model the use of a given filter in the time domain of the subband domain. The virtual alias-free performance that is especially important in HRTF-related applications is made possible as the phase properties of a filter in the time domain are (almost) perfectly transferred to the subband domain. Examples showing this will be outlined below.

Among the advantages of carrying out the second aspect of the invention are, in particular, the significant gain achieved by the computational efficiency. The complex modulated filter banks in the design of the new filter converter produce several subband signals with complex values ​​such as. the intermediate filter definition signal where each of the complex valued subband signals is shorter than the pulse response signal indicating the amplitude / frequency filter characteristic in the time domain. Accordingly, the filter converter produces an intermediate filter definition signal which comprises output signals from the complex modulated filter bank with its multiple short subband signal with complex values ​​that not only enable a fast and efficient and parallel calculation of a time domain's input signal to obtain a time domain's output signal of a time domain's output signal. design of a filter apparatus but also enables a quick and efficient and parallel calculation of the intermediate filter definition signal itself. Compared to a direct application of the pulse response signal indicating the amplitude / frequency filter characteristic of the time domain by wrapping the pulse response signal with the time domain input signal, the use of an embodiment of a new filter converter according to the second aspect of the invention makes it possible to achieve a simplified and more efficient calculation. leads to an audibly improved result compared to the more complex wrapping method.

Furthermore, an embodiment of the new filter converter also offers the advantage of a significantly improved flexibility in the possible filter characteristics used in the subband domain. Since arbitrary filter characteristics can be transmitted from the time domain to the subband domain using an embodiment of the new filter converter, tremendous flexibility is introduced into the audio signal processing and manipulation. Eg. For example, an embodiment of the new filter converter can provide an intermediate filter definition signal corresponding to an individually altered filter characteristic of an HRTF-related filter. In the area of ​​HRTF, this could allow individually to modify HRTF filters as needed and after a person's hearing. Furthermore, the source position as well as the listening position in relation to each other and in relation to a (simulated or calculated) environment (eg a concert hall, an open room, a stadium) can be used. This gives a great advantage to the listener with greater flexibility in relation to the acoustic conditions. Accordingly, an embodiment of the new filter converter provides the ability to virtually switch from a stadium to a concert hall or open space without having to transmit audio signals between the time domain, the subband domain and / or the frequency domain. By using an embodiment with the new filter converter, all these manipulations of the audio signal can be carried out within the subband domain of a very high quality that cannot be comprehensively separated from a signal processing in the time domain, but which offers a huge computational efficiency improvement.

This flexibility is not just limited to switching from one environment to another, e.g. switch from a stadium to a concert hall and vice versa. An embodiment of the new filter converter provides an opportunity to change the filter characteristics of several intermediate filters in a quasi-continuous manner. An application in the field of HRTF is an application of an embodiment of the filter converter and / or filter apparatus in a head tracking application where e.g. The position of the listener in relation to different sound sources varies in a quasi-continuous manner. Possible uses include e.g. simulations and computer games with very high quality.

Another advantage of the performance of a filter converter is that the use of an embodiment of a filter converter becomes more efficient in memory usage since a pulse response signal provided by the complex modulated filter bank of the filter converter is typically a truly valued signal while the mehom filter definition signal is a complex valued signal. about the same total length. As a result, the storage of the pulse response signals compared to the mehom filter definition signals (or the filter outputs of the mehom filters) saves in the order of about 2. Due to the possibility of a fast and efficient parallel calculation, especially in the area of ​​memory-sensitive applications that include a large parameter space relative to the possible pulses. response signals, this represents a significant advantage.

In one embodiment of the new filter converter, the filter converter is provided with a filter definition signal which can e.g. include filter outputs in a digital filter in the time domain or by a transfer function in the frequency domain which may include the amplitude / frequency characteristic and / or the phase / frequency characteristic of a filter. In this case, an embodiment of the filter converter further comprises a pulse response signal generator which provides the current pulse response signal indicating the resulting amplitude / frequency filter characteristic in the time domain of the complex modulated filter bank of the filter converter. Accordingly, the use of a pulse response signal generator in some embodiments of the new filter converter may offer even greater flexibility in delivering the mehom filter definition signal since not only the pulse response signals in the form of discrete time signals can be supplied to an embodiment of the filter converter but also the filter outputs or the frequency outputs of the filter converter. the time domain can be transferred to the subband domain by a suitable embodiment of a filter converter.

Brief description of the figures

The invention will be described in more detail below with reference to the drawings, in which:

Fig. 1a shows the processing of a digital audio signal by means of subband filtering in a system comprising a filter converter and a filter apparatus; 1b shows a possible solution for a complex analysis bank, fig. 1c shows a possible solution for a complex synthesis filter bank; FIG. 1d shows a possible solution for a complex synthesis filter bank, fig. 1c shows an interaction between an embodiment of a filter converter with several intermediate filters of an embodiment of a filter apparatus; FIG. 2 shows the processing of a digital audio signal by means of direct form filtering; 3 shows a preferred embodiment of a system with a filter converter; FIG. 4 shows a given filter pulse response; FIG. 5 shows a pulse response obtained by complex strength adjustment of partial bands; FIG. 6 shows the size response of a given filter; FIG. 7 shows the size response of a filter obtained by complex strength adjustment of partial bands; FIG. 8 compares the performance of the invention with a complex strength adjustment of subbands; Fig. 9 shows a preferred embodiment of a filter apparatus comprising an optional embodiment of a filter converter and other components; 10 shows a filter characteristic together with several frequency bands for different sub-bands; and FIG. 11 shows a preferred embodiment of a filter converter.

Detailed description

The embodiments described below further illustrate the principles of the invention for efficient filtration with a complex modulated filter bank. It will be appreciated that modifications and variations of the devices and details described herein will be apparent to one skilled in the art. Accordingly, it is intended only to limit the description to the scope of the claimed claims and not to the specific details presented in the description and explanation of the embodiments herein.

In the following, objects with the same or similar functional properties are denoted by the same reference character. Unless otherwise indicated, the description of objects having similar or similar functional properties may be exchanged relative to one another.

Fig. 1a shows in the form of a system comprising the design of both a filter apparatus and a filter converter, the processing of a digital audio signal by means of subband filtering according to the invention. This signal path may e.g. be part of a spatial audio reproduction system where the input is a received audio channel and the output is a component of a signal played back to the right ear. The input signal (digital audio signal or time domain input signal) is analyzed by the complex assay bank 101 by filtering with a set of L-analysis filters followed by downsampling of a factor L, where L is a positive integer, preferably greater than 1. Typically, the factor is L a power of 2, preferably L = 64. The analysis filters are usually obtained by a complex modulation of a prototype filter p (y), where v is a positive integer indicating an index in a group of data or an index of a value in a signal not sampled by a factor L. The signal from the filter bank consists of L subband signals processed by a subband filtering 102. This subband filtering consists of a combination of manipulations, e.g. subband gain adjustment according to the received control data and use of final pulse response filters applied separately in each subband. The filter outlets of the subband filters are retrieved from a (new) filter converter 104 as an embodiment of a filter converter having an input described by direct form filter outputs, a frequency domain description, or a pulse response (signal). The complex synthesis bank 103 reconstructs an output signal by collecting with a factor L, filtering L synthesis filters, summing all the results and extending the real part. The summation of all the results and the calculation of the real part can also be switched in their order as described in connection with FIG. lc and ld.

Fig. 1b shows a complex analysis bank 101 in detail. The complex assay bank 101 comprises multiple L intermediate analysis filters 120 for each subband transmitted by the complex assay bank 101. More precisely, each of the intermediate analysis filters 120 is connected in parallel to a node 130 to which the time domain input signal to be processed is provided. Each of the intermediate analysis filters 120 is adapted for filtering the input signal of the complex assay bank 101 relative to a center frequency of each subband. According to the center frequency of the different subbands, each subband is labeled with a subband index or index n, where n is a non-negative integer, typically in the range 0 to L1. The intermediate analysis filters 120 of the complex analysis bank 101 can be derived from a prototype filter p (y) by a complex modulation according to the subband index n of the subband on which the intermediate analysis filter 120 is applied. More details on the complex modulation of a prototype filter are explained below.

Either directly by means of the intermediate analysis filters 120 or by any down sampler 140 (shown dashed in Fig. 1b), the sampling frequency of the signal from the intermediate analysis bank 120 is reduced by a factor L. As mentioned earlier, the down samples 140 are applied to the output of each subband signal of the corresponding intermediate analysis filters 120. optional as, depending on the specific implementation, the downsampling can also be performed in the framework of the intermediate analysis filters 120. In principle, down-sampling of the signal from the intermediate analysis filters 120 is not required. In any case, the presence of explicit or implicit down samplers 140 is a preferred choice since the amount of data provided by complex assay bank 101 may alternatively be lifted by a factor of L and result in significant redundancy of data.

Fig. 1c shows a possible solution for a complex synthesis bank 103. The complex synthesis bank 103 comprises L intermediate synthesis filters to which L subband signals from subband filtering 102 are supplied. Depending on the specific implementation of complex synthesis bank 103 before filtering in the framework of intermediate synthesis filters 150, the subband signals are sampled by the L collector 160 which reconstructs the sampled frequency of the subband signals by increasing the sampling frequency by factor L. In other words, it reconstructs any sample sampler 160 or reformats the subband signals provided to sampler 160 such that the information in each of the subband signals is retained while the sampling frequency is increased by a factor L. As already explained in connection with FIG. 1, whatever the sampler 160 becomes optional components as the sampling can also be performed in the framework of the intermediate synthesis filters 150. Accordingly, the step of sampling the subband signals performed by the sampler 160 can be simultaneously processed in the framework of the intermediate synthesis filters 150. However, if the downsamplers 190 are neither explicitly nor implicitly implemented, the samplers 160 need not be explicitly or implicitly implemented.

The intermediate synthesizer filters 150 are coupled via an output to an adder 170 which sums up the filtered subband signals from the L intermediate synthesizers 150. The adder 170 is further coupled to a real subtractor 180 which extracts or forms a properly valued signal or rather a (really valued) time domain output. the complex valued signal provided by the adding unit 170. The real subtractor 180 can perform this task e.g. by extracting the real portion of a complex valued signal provided by the adding unit 170 and by calculating the absolute value of the complex valued signal provided by the adding unit 170 or by another method forming a truly valued output signal based on a complex valued input signal. In the case of the system shown in FIG. 1a, the signal from the real sub-extractor 180 is the output of the time domain according to the embodiment of the new filter apparatus.

The second possible solution for a complex synthesis bank 103 shown in FIG. Id differs from the first possible solution of FIG. 1c only at the actual sub-extractors 180 and the adder unit 170. More precisely, the signals from the intermediate synthesizer filters 150 are coupled separately from each subband to a real sub extractor 180 which extracts or forms a truly valued signal based on the complex valued signal from the intermediate synthesis filters 150. The actual sub-extractor 180 is then coupled to the add-on unit 170 which summarizes L really valued signals derived from L-filtered subband signals to form the actual valued output signal supplied by the add-on unit 170 as in the case of the system shown in FIG. 1a, is the output of the time domain.

Fig. 1c shows the partial band filtering 102 and its interaction with the filter converter 104 in detail. Subband filtering 102 comprises several intermediate filters 190, where an intermediate filter 190 is provided for each complex valued subband signal supplied to subband filtering 102. Accordingly, subband filtering 102 L comprises intermediate filters 190.

The filter converter 104 is coupled to each of the intermediate filters 190. As a result, the filter converter 104 can provide filter outlets for each of the intermediate filters 190 of the subband filtering 102. More details on the filtering performed by the intermediate filters 190 will be explained below. Accordingly, the filter outlets provided by the various intermediate filters 190 and transmitted by the filter converter 104 form the intermediate filter's defining signal.

Furthermore, it will be appreciated that the embodiments, solutions and implementations may include additional delays and / or any delays to delay some of the signals or a subset of signals that have been omitted in FIG. la-le for simplicity's sake. Also in FIG. 2-11, any delays have been omitted for the sake of simplicity. Regardless, delays or delay units may be included in the elements shown (e.g. filters) or added as any elements in all designs depending on their specific implementation.

FIG. 2 shows the processing of a digital audio signal by direct form filtering 201. If the same filter is provided as input to the filter converter 104 of FIG. 1 and the direct filtering 201, it becomes a design goal of the filter converter 104 that the digital audio signal from 103 becomes perceptually (or audibly) indistinguishable from the digital audio signal from the direct filtering 201 if the digital audio signals to the complex assay bank 101 and the direct filtering 201 are identical and the treatment in the direct filtration 102 consists of pure stationary subband filtering.

In the embodiment of the system shown in FIG. 1 to FIG. 1c, the filter signal of filter converter 104 is given as a filter definition signal such as e.g. may include filter outputs from a corresponding time domain filter, a frequency domain description (amplitude / frequency characteristic and / or phase / frequency characteristic), or a pulse response signal of the current filter.

In the case of direct filtering 201, the same filter definition signal can in principle be used. Depending on the specific implementation and filter definition signal, the filtering can be carried out using direct filter outputs in the framework of a digital filter by means of a discrete Fourier transform together with a transfer function or other frequency domain description or by a envelope with the pulse response signal.

FIG. 3 shows a preferred embodiment of a filter converter 104 according to the invention as an embodiment of a filter converter. The filter is assumed to be given by its pulse response. By considering this pulse response as a discrete time signal, it is analyzed by an L-band complex analysis (filter) bank 301. The resulting subband signal then becomes the exact pulse responses of the filters used separately in each subband in subband filtering 102. In the preferred embodiment shown in FIG. 3, the filter definition signal is delivered to the filter converter 104 and its complex analysis bank or complex analysis filter bank 301 the pulse response signal indicating the amplitude / frequency characteristic of a filter transmitted to the subband domain. Accordingly, the signal from the complex analysis (filter) bank 301 of each of the L-band bands represents the pulse response of the intermediate filters included in the sub-band filtering 102.

The complex assay bank 301 is derived in principle from the assay bank 101, but it has a different type of prototype filter and a slightly different modulation structure, the details of which will be outlined in the description below. The same fast algorithms used for the implementation of the complex assay bank 101 can be used again for the complex assay bank 301 and result in a very fast and very efficient conversion process.

Furthermore, the length of the prototype filter q (y) can be constructed only to be a fraction of the length of the prototype filter p (v). Also, because of the downsampling by a factor L, the length of the subband filters is a factor of L less than the sum of the lengths of the given time domain filter and the prototype filter q (y). The calculation work is then reduced compared to direct mold filtration 201 by about a factor of L / 4. The offset factor of 4 is due to the replacement of the actual filtration with a complex filtration. Another offset is the computational cost of the complex analysis and the synthesis banks 101 and 103. For efficient implementations, this cost is comparable to the cost of fairly short FIR filters and consequently negligible, as mentioned earlier. Furthermore, this offset of reduction in computational costs will not be found in systems that already use these two filter banks 101 and 103.

Fig. 4 shows an example of a given filter pulse response 400. It consists of 192 (= 64 · 3) non-zero outputs. In other words, the pulse response 400 of FIG. 4,192 non-vanishing values.

In this application, a non-vanishing withdrawal or value is a withdrawal or value that is ideally not equal to zero. Due to implementation constraints in the framework of this application, whatever a non-vanishing value or withdrawal is, a truly valued or complexly valued withdrawal or a value with an absolute value greater than a certain threshold, e.g. 10 '⁵ or 2 ~^s, where γ is a positive integer depending on the requirements for a concrete implementation. In digital systems, this threshold is preferably defined in the binary system (base 2), where integers have a specific value depending on the implementation specification. Typically, the value is 4, 5, 6, 7, 8, 10, 12, 14, 16 or 32.

The pulse response 400 in the system of FIG. 1 cannot be distinguished from this given pulse response in resolution of the image in a case where an L = 64 band filter bank with a prototype filter of length 640 (= 64-10) is used and a prototype filter of length 192 (= 64 · 3) is used for the filter converter 104 of FIG. 3. The corresponding middle band filters have only 5 (= 3 + 3-1) outputs each as explained below.

FIG. 5 shows the pulse response 410 of the system of FIG. 1 with a 64-band filter bank in a special case corresponding to the current use of envelope alignment and leveling. In this case, the subband filters or rather the intermediate filters 190 are all only of an outlet, so that a constant, complex gain is applied to each subband. For each subband, corresponding gain is chosen to be equal to the complex frequency response of the filter of FIG. 4 evaluated at the center frequency of the particular subband. As can be seen from the result, there are several pre-echo cases and there will be a significant perceptual difference between the use of this filter response compared to the target pulse response 400 in FIG. 4.

FIG. 6 shows the size response 420 of the filter of FIG. 4. The frequency scale of FIG. 6 is adjusted to the resolution of a 64-band filter bank (L = 64).

FIG. 7 shows the size response 430 of the filter below the pulse response 410 shown in FIG. 5. As can be seen, the use of only one gain per subband leads to poor approximation to the desired frequency response. The main reason for this is the rapid variation of the target phase spectrum. In fact, this method of the present technique is better suited for modeling linear phase responses.

FIG. 8 finally compares the performance of an embodiment of the invention with the performance of the current technique of complex strength adjustment of partial bands. The dashed curve is a representation of the target size response 420 of FIG. 6. The dashed curve 440 is the magnitude response of the difference between the complex frequency responses of the target filter and its approximation by the current method. The solid-state curve 450 is the magnitude response of the difference between the complex frequency responses of the target filter and its approximation by the method of the invention wherein the parameters as mentioned in the description of FIG. 4. As will be seen, the error of the method of the present technique is small only at the 64 midpoints of the filter bank subband while the new method results in an approximation quality in the 50 dB range. It should be emphasized that this is also the level of performance being measured compared to the signal from the new system to the output of the reference system for any input signal.

By comparison, the two curves 440 and 450 in FIG. 8 shows, the embodiment of the new filter apparatus, an embodiment of the filter converter and a system comprising both embodiments gives a considerable advantage in quality in the manipulation of an input signal. The significant difference in the quality of the filtering (or manipulation) of the input signal outlined above is a result of at least one of the intermediate filters 190 having a pulse response of two or Here non-vanishing values. In other words, at least one of the intermediate filters 190 comprises at least two non-vanishing filter outlets. Furthermore, it is important to note that the number of subbands L processed by one embodiment of a filter apparatus is greater or at least equal to 2. In any case, the number of sub-bands L becomes significantly smaller than the number of frequency bands required for comparable quality in a Fourier transform-based filtering compared to a filter mainly described with an amplitude / frequency characteristic and / or a phase / frequency characteristic as the transfer function of the filter.

Because the impulse response of the intermediate filters 190 is substantially shorter than the pulse response of the underlying filter characteristic in the time domain, the calculations for each subband can be performed significantly faster. Since the various subband signals can be processed independently, both an embodiment of the filter apparatus and an embodiment of the filter converter 104 can process the respective input signals very efficiently in a fast and parallel manner. Accordingly, the processing of both a digital audio input signal as well as a pulse response indicating a filter characteristic can be performed very efficiently in a parallel manner. As mentioned earlier, an embodiment of a new filter apparatus as well as an embodiment of a new filter converter combine the advantages of both a direct processing of audio signals in the time domain which led to very high quality and the use of a combination of a Fourier transform together with a transmission function in the frequency domain. high efficiency since each frequency band is only multiplied by a (complex or really valued) output in the processing of signal filtering.

On the other hand, the disadvantages of both the pure processing of the input signals in the time domain, which leads to enormous computational work and the Fourier transformation, result in a significant reduction and suppression to a level in that the signal from an embodiment of a filter device cannot be distinguished from the quality of a filter device. direct processing in the time domain.

These two advantages provide great flexibility for filtering the digital signals with varying filter characteristics. This is especially important in the area of ​​HRTF as HRTF-related filters often have very long pulse responses. Accordingly, an embodiment of the new filter apparatus comprising a complex assay filter bank 101, multiple intermediate filters 190 in the subband filtering 102, and a complex synthesis filter bank 103, especially in the field of HRTF-related applications, provides significant computational advantages due to the possible parallel processing of subband signals.

The design of a filter converter and the design of systems comprising both a filter apparatus and a filter converter further provide the advantage that filters can be specially adapted to specific environments, parameters or other specific needs for the particular application.

Particularly in the case of HRTF-related applications, such an embodiment of a system can be used for head tracking applications where multiple sources of noise and sound as well as the position of the listener may vary over time. Such an embodiment of a system comprises a filter apparatus and filter converter which consequently provides a very effective and flexible way of presenting a sound impression in a three-dimensional arrangement of sound sources in relation to a varying position and arrangement of a hypothetical listener via headphones or other head-related sound systems (stereo sound systems ).

As this latest example demonstrates, an embodiment of the new filter apparatus together with a new filter converter not only offers a very efficient audio manipulation system of excellent quality but also a very flexible way of introducing changed sound impressions, in an efficient manner.

Complex modulated filter banks

In the following, Ζ (ω) = ^ 2 ”æz (v) exp (-zvcy) is the discrete-time Fourier transform of a discrete-time signal z (y). As before, v is an integer indicating an index or time index of a time signal while ω = 2 π - / is the circular frequency associated with the frequency / while π is the circular number (π = 3.1415926 ...) and i = j = YX is the thought unit.

The complex exponentially modulated L-band filter bank is defined by a truly valued prototype filter p (y) of finite length. For the calculations below, it will be assumed by extension by zeros that the prototype filter is defined for all integers v. Given a truly valued, discrete time signal x (v), the analysis filter bank 101, as already explained, uses the complex, modulated prototype filters followed by downsampling by factor L for sending the subband signals, c_n(k) = ^ x (y + kL) p (y) exp

<img file="NO342467B1_D0002.tif" />

(1) for each subband index n = 0.1, ..., Ll, and integer time index k. The time index k differs from the time index v in that k refers to the downsampled signals while integers v creep to full sample frequency signals.

Given complex, valued subband signals d_n(k), the synthesis filter bank 103 uses filtering followed by sampling with a factor L and a fair value extraction to send the real valued signals already explained to obtain the output signal oo Ll (rr 1 y (v) = Re &lt;2Λ Σ ^ &lt; 7 "(Æ) p (v-ÆL) exp i - (n + -) (v-kL + ys) k = -ec η = Π \ L 2 (2)

In Equations (1) and (2), Θ and ψ (constant) represent phase factors for filtering the truly valued discrete time signal x (v) to complex valued subband signal and to reconstruct really valued output signals y (v) from subband signal with complex values. d_n(K). It is well known that a prototype filter and solid phase factors Θ and ψ can be chosen to give a perfect reconstruction y (v) = x (v) in the case where d_n(k) = c_n(k), i.e. when the subband signals are unchanged. In practice, the perfect reconstruction feature will be true up to a delay (and / or a character change), but in the calculations that follow, this detail will be ignored using an acausal prototype filter. The invention applies to quasi-QMF type construction as described in PCT / SE02 / 00626 Aliasing reduction using complex exponential modulated filter banks. Here, the prototype filter is symmetric p (-v) = p (v) and its discrete-time Fourier transform Ρ (ω) disappears substantially outside the interval | æ »| &lt;æ / L. The perfect reconstruction is also replaced by an almost perfect reconstruction feature. For the derivation that follows, it will be assumed for simplicity that both a perfect reconstruction holds and that Ρ (ω) = 0 for π IL &lt;\ a \ &lt;π. Furthermore, these factors are assumed to fulfill the ratio that ψ- Equals an integer multiple of 4L.

In a critically sampled filter bank, the change of the subband signals before synthesis usually results in the introduction of foreign elements. This is solved here by a factor two oversampling and uses complex valued signals. Although the total sampling rate of the subband signals is identical to the sampling rate of the discrete time input signal, the input signal is truly appreciated and the subband samples are complexly valued. As mentioned below, the absence of foreign elements opens the door for efficient time invariance signal processing.

Part-band filtering in a complex modulated filter bank

Given the modification of the subband filtering 102 of each subband signal obtained by filtering the analysis samples c_n(k) from the complex assay bank 101 with a filter with pulse response g "(k) before the synthesis (2) performed by the complex synthesis (filter) bank 103, becomes:

= (3) l

Elementary calculations show that given assumptions about the frequency response of the prototype filter, the resulting weight of the reconstructed time signal becomes a discrete time filtering

Y (&lt;y) = G (&lt;y) X (&lt;y) (4) where

G (ro) = ^ G „(Lrø) n = -L p \ ^ω~ τ ^ (5)

Here's G_n (ω) = Σ, « _n(k) exp (-ico) the discrete-time Fourier transform of the filter used in the subband n for n> 0 and

G „(ω) = G__x__n (-ω) * for n <Q. (6) where * denotes complex conjugation. Note here that the special case G "(&>) = 1 results in a G (<w) = 1 in (5) due to the assumed special design of the prototype p (yj which implies

L, -1

7 (7)

Another interesting case is G „(&>) = exp (- / &>) leading to G (a) = exp (-iLa) such that y (v) = x (v-Lj.

Approximation of a given filter response by subband filtering is Let Η (ώ) be a given filter (eg, transform function) with a truly valued pulse response h (yy. This data is considered to be sent to filter converter 104. Considering (5) and (7) ), a trivial choice for the subband filters leading to the desired response Ο (ω) = Η (ω) is given by

Ο_η(ω) = Η (ω! L), for | rø-Æ (n + l / 2) | &lt;π, (8)

The disadvantage of this formula is that although Η (ώ) is a smooth function of ω, the periodic segment thereof as defined by (8) will show jumps and the pulse response of the subband filters will be unnecessarily long. The current use of the complex quasiQMF bank for leveling or wrapping adjustment consists of using a single reinforcement g_n in each subband leading to the transfer function ο (®) = ​​Σ «.

n = -L for n = 0.1, ..., Ll (10) (9) with the extension g_n = -gt₁For n &lt;0 as defined in (6). Considering (7), it will be achieved

G_n(coj = H \ ^ (n + ^ and the transfer function is interpolated between these frequencies. For the target filter responses Η (ω) which vary slowly as a function of the frequency ω, a first method of approximating the filter is thus obtained by selecting

)] (11)

An example of the resulting quality of this method is given in FIG. 5 and 7.

According to an embodiment of the invention, a filter converter or filter converter 104 is used to learn how to convert the filter (as defined by its io pulse response) h (y) to intermediate band filters 190 by means of the second analysis filter bank.

301 using real valued prototype filter q (y), g „(k) = ^ h (v + kL) q (v) exp \ -ίγ (η + ^) ν f.

When it comes to Fourier transforms, this will be read as

-.....-; ςή + (12) (13)

The advantage of this method is that a given filter h (yj can be effectively transformed into intermediate band filter responses. If q (y) has Kq-L outputs, a time domain filter h (yj of Kh-L outputs is converted to subband domain filters (12) where Kh +

Kq - 1 outlet, where Kh and Kq are positive integers. In the case of the example numbers given in connection with the description of FIG. 4, Kh and Kq are equal to 3 and with a prototype filter length and a pulse response corresponding to a length of 3 · 64 = 192 (L = 64) each. Accordingly, each interband band filter 190 has a pulse response length of only 3 + 3-1 = 5 outputs each.

Design of the prototype filter for the filter converter Insertion of (13) into (5) gives

LL 2

Accordingly, to obtain the condition of G ((o) = Η (ώ) is that i = 0 VL) n = -L ^^ 1 (2π π 1?

XtU '+' U '.

= 4/], (14) (15)

where &lt;5 [Z] = 1 for l = 0 and &lt;5 [Z] = 0 for l * 0. A simple solution to (15) is given by the brick wall filter

Q (X =

L,

0, V for \ a> \ <π / L; forx; IL <Icyl <π

This prototype filter corresponds to selection (8) and the disadvantage of infinite and slow attenuating pulse response q (y). Instead, the invention seeks to solve (15) approximately (e.g., in the least squares sense) with a finite pulse response filter q (y). The time domain corresponding to (15) is the system of linear equations for n = 0.1, ... Ll and for all integers k,

CO 1 y, p₇ (n + vL - 2kL) q (n + vL) = —L> '[Æ |, v— m 2L (16) where

L₂(^v) = Z ^ (0W + v) (17) l = -co is the autocorrelation of p (v). For a given support length of the system of linear equations (16), the least squares significance for a prototype filter q (y) can be solved. It is desirable to use a support that is significantly shorter than in the original filter bank prototype filter p (y) and in the case where the linear system (16) becomes overdetermined. A given quality of the approximation can also be exchanged for other desirable properties via an overall optimization. An example of such a property is a frequency of the low-pass type ζ) (ω).

In the following, the determination of a multi-slot QMF rendering (the subband domain) of the HRTF filters is described. The filter conversion from the time domain to the complex QMF subband domain is performed by an FIR filter in the filter converter 104 of FIG. let. More precisely, the following description outlines a method for implementing a given FIR filter h (v) of length Nh in the complex QMF subband domain. The principles of the operation are shown in FIG. in the case of a system which also comprises an embodiment of a new filter apparatus.

The subband filtering itself is performed by a set or more intermediate filters 190 within the subband filtering 102. More precisely, the subband filtering consists of the separate application of a complex valued FIR intermediate filter g "(Z) where each QMF subband with an index n = 0.1, .. In other words, in the following description, embodiments with L = 64 different subband signals will be referred to. However, this specific number of subband signals is not significant and the corresponding equations will also be given in a more general form.

One of the most important components of the system shown in FIG. la, is the filter converter

104 which converts the given time domain FIR filter h (y) to the complex subband domain filters g_n(T). Filter converter 104 comprises a complex assay bank 304 corresponding to QMF assay bank 101. The prototype filter of the complex assay filter bank 301 of the filter converter 104 q (y) of length 192 (= 3 · 64) for the specific case where L = 64 subband signals are generated by resolving at least , in the least squares sense, the overdetermined system of the equation (16). The filter coefficients q (v) or rather the conditions they fulfill will be described in detail for the case where L = 64 subband signals come later.

To be more accurate in the mathematical description, an extension of zeros in the time domain's FIR filter is defined by h (vv = &lt;

h (v (v = 0, l, .., N)_h0, otherwise

-1, (18)

The resulting mid-band domain filters are based on Equation (12) and can be expressed in the general case as „„ (/) = ζΛ (ν + · · (/ - /₀)) · ^ (Ν) · εχρ ^ -Χ ^ + | ^ (ν-ν_θ(19) where lo and vo are delays, l is an integer indicating an index of the filter outlets and / V_?(= Nq) is the length of the pulse response of the prototype filter q (y).

It should be noted that in the framework of the present application, under an equation based on an equation, it should be understood to introduce additional delays (see lo and vo) factors, multiple coefficients and the introduction of a window function or other simple function.

In the case where L = 64, the expression of the subband domain filters or intermediate filters becomes 190 g + (/) = 22^{v +}64 - (/ - 2)) - ^ (v) -^exp ^ -z ^ - ^ + ^ (v-95) J (20)

These subband filters have a length L_q = Kh + 2, there

VM (21) and Nh are the lengths of the pulse response h (y) of the filter characteristics to be transmitted to the subband domain.

In this case, integers n = 0, 1, ..., 63 are again the index of a subband and / = 0,

1, ..., (Kh + l) is an integer indicating the output of the resulting intermediate filters 190.

The additional addition of (-2) in Equation (20) compared to Equation (12) is because Equation (12) is developed without taking into account the causality of the filters.

Real implementations will always introduce delays. Accordingly, depending on the specific implementation, additional delay units or delays can be implemented in the embodiments of FIG. 1a and 1c. 2-11 which have been omitted for the sake of simplicity in the above figures.

As described earlier, in many cases the system of linear equations (16) is overestimated. Accordingly, it cannot be solved or approximated in the least square sense relative to the prototype filter coefficients q (y). Solution of the system of linear equations (16) in the least square sense leads to the filter output of the prototype filter q (y) to fulfill the following relation for the integers v from 0 to 191:

-0.204 &lt;q [0] &lt;-0.202 -0.199 &lt;q [l] &lt;-0.197 -0.194 &lt;q [2] &lt;-0.192 -0.189 &lt;q [3] &lt;-0.187 -0.183 &lt;q [4] &lt;0.181 -0.187 &lt;q [5] &lt;-0.176 -0.172 &lt;q [6] &lt;-0.170 -0.166 &lt;q [7] &lt;-0.164 -0.160 &lt;q [8] &lt;-0.158 -0.154 &lt;q [9] &lt;-0.152 -0.148 &lt;q [10] &lt;-0.146 -0.142 &lt;q [ll] &lt;-0.140 -0.135 &lt;q [12] &lt;- 0.133 -0.129 &lt;q [13] &lt;-0.127 -0.122 &lt;q [14] &lt;-0.120 -0.116 &lt;q [15] &lt;- 0.114 -0.109 &lt;q [16] &lt;-0.107 -0.102 &lt;q [17] &lt;-0.100 -0.096 &lt;q [18] &lt;-0.094 -0.089 &lt;q [19] &lt;-0.087 -0.082 &lt;q [20] &lt;-0.080 -0.075 &lt;q [21] &lt;-0.073 -0.068 &lt;q [22] &lt; -0,066 -0,061 &lt;q [23] &lt;-0,059 -0,054 &lt;q [24] &lt;-0,052 -0,046 &lt;q [25] &lt;-0,044 -0,039 &lt;q [26] &lt;-0,037 -0,032 &lt;q [27] &lt; -0.030 -0.024 &lt;q [28] &lt;-0.022

-0.017 &lt;q [29] &lt;-0.015 -0.009 &lt;q [30] &lt;-0.007 -0.002 &lt;q [31] &lt;0.000 0.006 &lt;q [32] &lt;0.008 0.014 &lt;q [33] &lt;0.016 0.021 &lt;q [34 ] &lt;0.023 0.029 &lt;q [35] &lt;0.031 0.037 &lt;q [36] &lt;0.039 0.045 &lt;q [37] &lt;0.047 0.054 &lt;q [38] &lt;0.056 0.062 &lt;q [39] &lt;0.064 0.070 &lt;q [40] &lt; 0.072 0.079 &lt;q [41] &lt;0.081 0.087 &lt;q [42] &lt;0.089 0.096 &lt;q [43] &lt;0.098 0.105 &lt;q [44] &lt;0.107 0.113 &lt;q [45] &lt;0.115 0.122 &lt;q [46] &lt;0.124 0.132 &lt;q [47] &lt;0.134 0.141 &lt;q [48] &lt;0.143 0.150 &lt;q [49] &lt;0.152 0.160 &lt;q [50] &lt;0.162 0.170 &lt;qt51] &lt;0.172 0.180 &lt;q [52] &lt; 0.182 0.190 &lt;q [53] &lt;0.192 0.200 &lt;q [54] &lt;0.202 0.210 &lt;q [55] &lt;0.212 0.221 &lt;q [56] &lt;0.223 0.232 &lt;q [57] &lt;0.234 0.243 &lt;q [58] &lt;0.245 0.254 &lt;q [59] &lt;0.256 0.266 &lt;q [60] &lt;0.268 0.278 &lt;q [61] &lt;0.280 0.290 &lt;q [62] &lt;0.292 0.303 &lt;q [63] &lt;0.305 0.902 &lt;q [64] &lt;0.904 0.909 &lt;q [65] &lt;0.911 0.917 &lt;q [66] &lt;0.919 0.924 &lt;q [67] &lt;0.926

0.930 &lt;q [68] &lt;0.932 0.936 &lt;q [69] &lt;0.938 0.942 &lt;q [70] &lt;0.944 0.947 &lt;q [71] &lt;0.949 0.952 &lt;q [72] &lt;0.954 0.957 &lt;q [73] &lt;0.959 0.961 &lt; q [74] &lt;0.963 0.965 &lt;q [75] &lt;0.967 0.969 &lt;q [76] &lt;0.971 0.972 &lt;q [77] &lt;0.974 0.975 &lt;q [78] &lt;0.977 0.978 &lt;q [79] &lt;0.980 0.981 &lt;q [ 80] &lt;0.983 0.984 &lt;q [81] &lt;0.986 0.986 &lt;q [82] &lt;0.988 0.988 &lt;q [83] &lt;0.990 0.990 &lt;q [84] &lt;0.992 0.992 &lt;q [85] &lt;0.994 0.993 &lt;q [86] &lt;0.995 0.995 &lt;q [87] &lt;0.997 0.996 &lt;q [88] &lt;0.998 0.997 &lt;q [89] &lt;0.999 0.998 &lt;q [90] &lt;1,000 0.999 &lt;q [91] &lt;1.001 0.999 &lt;q [92] &lt;1.001 1,000 &lt;q [93] &lt;1.002 1,000 &lt;q [94] &lt;1.002 1,000 &lt;q [95] &lt;1.002 1,000 &lt;q [96] &lt;1.002 1,000 &lt; q [97] &lt;1,002 0.999 &lt;q [98] &lt;1,001 0.999 &lt;q [99] &lt;1.001 0.998 &lt;q [100] &lt;1,000 0.997 &lt;q [101] &lt;0.999 0.996 &lt;q [102] &lt;0.998 0.995 &lt;q [ 103] &lt;0.997 0.993 &lt;q [104] &lt;0.995 0.992 &lt;q [105] &lt;0.994 0.990 &lt;q [106] &lt;0.992

0.988 &lt;q [107] &lt;0.990 0.986 &lt;q [108] &lt;0.988 0.984 &lt;q [109] &lt;0.986 0.981 &lt;q [l 10] &lt;0.983 0.978 &lt;q [l 11] &lt;0.980 0.975 &lt;q [l 12] &lt; 0.977 0.972 &lt;q [l 13] &lt;0.974 0.969 &lt;q [l 14] &lt;0.971 0.965 &lt;q [l 15] &lt;0.967 0.961 &lt;q [l 16] &lt;0.963 0.957 &lt;q [l 17] &lt;0.959 0.952 &lt;q [ l 18] &lt;0.954 0.947 &lt;q [l 19] &lt;0.949 0.942 &lt;q [120] &lt;0.944 0.936 &lt;q [121] &lt;0.938 0.930 &lt;q [122] &lt;0.932 0.924 &lt;q [123] &lt;0.926 0.917 &lt;q [ 124] &lt;0.919 0.909 &lt;q [125] &lt;0.911 0.902 &lt;q [126] &lt;0.904 0.893 &lt;q [127] &lt;0.895 0.290 &lt;q [128] &lt;0.292 0.278 &lt;q [129] &lt;0.280 0.266 &lt;q [130] &lt;0.268 0.254 &lt;q [131] &lt;0.256 0.243 &lt;q [132] &lt;0.245 0.232 &lt;q [133] &lt;0.234 0.221 &lt;q [134] &lt;0.223 0.210 &lt;q [135] &lt;0.212 0.200 &lt;q [136] &lt;0.202 0.190 &lt;q [137] &lt;0.192 0.180 &lt;q [138] &lt;0.182 0.170 &lt;q [139] &lt;0.172 0.160 &lt;q [140] &lt;0.162 0.150 &lt;q [141] &lt;0.152 0.141 &lt;q [142 ] &lt;0.143 0.132 &lt;q [143] &lt;0.134 0.122 &lt;q [144] &lt;0.124 0.113 &lt;q [145] &lt;0.115

0.105 &lt;q [146] &lt;0.107 0.096 &lt;q [147] &lt;0.098 0.087 &lt;q [148] &lt;0.089 0.079 &lt;q [149] &lt;0.081

0.070 &lt;q [150] &lt;0.072

0.062 &lt;q [151] &lt;0.064 0.054 &lt;q [152] &lt;0.0556 0.045 &lt;q [153] &lt;0.047 0.037 &lt;q [154] &lt;0.039 io 0.029 &lt;q [155] &lt;0.031 0.021 &lt;q [156] &lt;0.023 0.014 &lt;q [157] &lt;0.016 0.006 &lt;q [158] &lt;0.008

-0.002 &lt;q [159] &lt;0.000 is -0.009 &lt;q [160] &lt;-0.007

-0.017 &lt;q [161] &lt;-0.015 -0.024 &lt;q [162] &lt;-0.022 -0.032 &lt;q [163] &lt;-0.030 -0.039 &lt;q [164] &lt;-0.037

-0.046 &lt;q [165] &lt;-0.044

-0.054 &lt;q [166] &lt;-0.052 -0.061 &lt;q [167] &lt;-0.059 -0.068 &lt;q [168] &lt;-0.066 -0.075 &lt;q [169] &lt;-0.073

-0.082 &lt;q [170] &lt;-0.080

-0.089 &lt;q [171] &lt;-0.087 -0.096 &lt;q [172] &lt;-0.094 -0.102 &lt;q [173] &lt;-0.100 -0.109 &lt;q [174] &lt;-0.107 so -0.116 &lt;q [175] &lt;- 0.114 -0.122 &lt;q [176] &lt;-0.120 -0.129 &lt;q [177] &lt;-0.127 -0.135 &lt;q [178] &lt;-0.133 -0.142 &lt;q [179] &lt;-0.140

-0.148 &lt;q [180] &lt;-0.146

-0.154 &lt;q [181] &lt;- 0.152 -0.160 &lt;q [182] &lt;-0.158 -0.166 &lt;q [183] ​​&lt;-0.164 -0.172 &lt;q [184] &lt;-0.170

-0.178 &lt;q [185] &lt;-0.176

-0.183 &lt;q [186] &lt;- 0.181

-0.189 &lt;q [187] &lt;-0.187

-0.194 &lt;q [188] &lt;-0.192 &lt;-0.119 &lt;q [189] &lt;-0.197

-0.204 &lt;q [190] &lt;-0.202

-0.209 &lt;q [191] &lt;-0.207

To be more accurate, the filter coefficients q (v) obey the following conditions:

-0.20294 &lt;q [0] &lt;-0.20292

-0.19804 &lt;q [l] &lt;-0.19802

-0.1995 &lt;q [2] &lt;-0.19293

-0.18768 &lt;q [3] &lt;-0.18766 is -0.18226 &lt;q [4] &lt;-0.18224

-0.17668 &lt;q [5] &lt;-0.17666

-0.17097 &lt;q [6] &lt;-0.17095

-0.16514 &lt;q [7] &lt;-0.16512

-0.15919 &lt;q [8] &lt;-0.15917 &lt;-0.15313 &lt;q [9] &lt;0.15311

-0.14669 &lt;q [10] &lt;-0.14695

-0.14071 &lt;q [ll] &lt;- 0.14069

-0.13437 &lt;q [12] &lt;-0.13435

-0.12794 &lt;q [13] &lt;-0.12792 25 -0.12144 &lt;q [14] &lt;-0.12142

-0.11486 &lt;q [15] &lt;- 0.11484

-0.10821 &lt;q [16] &lt;- 0.10819

-0.10149 &lt;q [17] &lt;-0.10147

-0.09471 &lt;q [18] &lt;-0.09469 so -0.08786 &lt;q [19] &lt;-0.08784

-0.08095 &lt;q [20] &lt;-0.08093

-0.07397 &lt;q [21] &lt;-0.07395

-0.06694 &lt;q [22] &lt;-0.06692

-0.05984 &lt;q [23] &lt;-0.05982 35 -0.05269 &lt;q [24] &lt;-0.05267

-0.04547 &lt;q [25] &lt;-0.04545

-0.03819 &lt;q [26] &lt;-0.03817

-0.03085 &lt;q [27] &lt;-0.03083

-0.02345 &lt;q [28] &lt;-0.02343

-0.01598 &lt;q [29] &lt;-0.01596 -0.00845 &lt;q [30] &lt;-0.00843 -0.00084 &lt;q [31] &lt;-0.00082 0.00683 &lt;q [32] &lt;0.00685 0.01458 &lt;q [33] &lt;0.01460 0.02240 &lt;q [34] &lt;0.02242 0.03030 &lt;q [35] &lt;0.03032 0.03828 &lt;q [36] &lt;0 , 03830 0.04635 &lt;q [37] &lt;0.04637 0.05451 &lt;q [38] &lt;0.05453 0.06275 &lt;q [39] &lt;0.06277 0.07110 &lt;q [40] &lt;0.07112 0.07954 &lt;q [41] &lt;0.07956 0.08809 &lt;q [42] &lt;0.08811 0.09675 &lt;q [43] &lt;0.09677 0.10552 &lt;q [44] &lt;0.10554 0, 11442 &lt;q [45] &lt;0.11444 0.12344 &lt;q [46] &lt;0.12346 0.13259 &lt;q [47] &lt;0.13261 0.14189 &lt;q [48] &lt;0.14191 0.15132 &lt;q [49] &lt;0.15134 0.16091 &lt;q [50] &lt;0.16093 0.17066 &lt;q [51] &lt;0.17068 0.18058 &lt;q [52] &lt;0 , 18060 0.19067 &lt;q [53] &lt;0.19069 0.20095 &lt;q [54] &lt;0.20097 0.21143 &lt;q [55] &lt;0.21145 0.22211 &lt;q [56] &lt;0.22213 0.23300 &lt;q [57] &lt;0.23302 0.24412 &lt;q [58] &lt;0.24414 0.25549 &lt;q [59] &lt;0.25551 0.26711 &lt;q [60] &lt;0.26713 0, 27899 &lt;q [61] &lt;0.27901 0.29117 &lt;q [62] &lt;0.29119 0.30364 &lt;q [63] &lt;0.30366 0.90252 &lt;q [64] &lt;0.90254 0.91035 &lt; q [65] &lt;0.91037 0.91769 &lt;q [66] &lt;0.91771 0.92457 &lt;q [67] &lt;0.92459

0.93101 &lt;q [68] &lt;0.93103 0.93705 &lt;q [69] &lt;0.93707 0.94270 &lt;q [70] &lt;0.9272 0.94800 &lt;q [71] &lt;0.94802 0, 95295 &lt;q [72] &lt;0.95297 0.95758 &lt;q [73] &lt;0.95760 0.96190 &lt;q [74] &lt;0.96192 0.96593 &lt;q [75] &lt;0.96595 0.96968 &lt; q [76] &lt;0.96970 0.97317 &lt;q [77] &lt;0.97319.97641 &lt;q [78] &lt;0.97643 0.97940 &lt;q [79] &lt;0.97942 0.98217 &lt;q [ 80] &lt;0.98219 0.98472 &lt;q [81] &lt;0.98474 0.9870 &lt;q [82] &lt;0.9870 0.8919 &lt;q [83] &lt;0.98921 0.99113 &lt;q [84] &lt;0.99115 0.99288 &lt;q [85] &lt;0.99290 0.99444 &lt;q [86] &lt;0.99446 0.99583 &lt;q [87] &lt; 0.99585 0.99704 &lt;q [88] &lt;0.99706 0.99809 &lt;q [89] &lt;0.99811 0.99896 &lt;q [90] &lt;0.99898 0.99967 &lt;q [91] &lt;0, 99969 00023 &lt;q [92] &lt;1,00025 00062 &lt;q [93] &lt;1,00064 00086 &lt;q [94] &lt;1,00088 00093 &lt;q [95] &lt;1,00095 00086 &lt;q [96] &lt;1, 00088 1,00062 &lt;q [97] &lt;1,00064 1,00023 &lt;q [98] &lt;1,00025 0.99967 &lt;q [99] &lt;0.99969 0.99896 &lt;q [100] &lt;0.99898 0 , 99809 &lt;q [101] &lt;0.99811 0.99704 &lt;q [102] &lt;0.99706 0.99583 &lt;q [103] &lt;0.99585 0.99444 &lt;q [104] &lt;0.99446 0.99288 &lt;q [105] &lt;0.99290 0.99113 &lt;q [106] &lt;0.99115

0.98919 &lt;q [107] &lt;0.98921 0.98706 &lt;q [108] &lt;0.98708 0.98472 &lt;q [109] &lt;0.9847 0.98217 &lt;q [l 10] &lt;0.98219 0 , 97940 &lt;q [l 11] &lt;0.97942 0.97641 &lt;q [l 12] &lt;0.97643 0.97317 &lt;q [l 13] &lt;0.9319 0.96968 &lt;q [l 14] &lt;0, 96970 0.96593 &lt;q [l 15] &lt;0.96595 0.96190 &lt;q [l 16] &lt;0.96192 0.95758 &lt;q [l 17] &lt;0.95760 0.95295 &lt;q [l 18] &lt; 0.95297 0.94800 &lt;q [l 19] &lt;0.94802 0.94270 &lt;q [120] &lt;0.94272 0.93705 &lt;q [121] &lt;0.93707 0.93101 &lt;q [122] &lt;0 , 93103 0.92457 &lt;q [123] &lt;0.9459 0.91769 &lt;q [124] &lt;0.91771 0.91035 &lt;q [125] &lt;0.91037 0.90252 &lt;q [126] &lt;0.90254 0.89416 &lt;q [127] &lt;0.89418 0.29117 &lt;q [128] &lt;0.29119 0.27899 &lt;q [129] &lt;0.27901 0.26711 &lt;q [130] &lt;0.26713 0.25549 &lt;q [131] &lt;0.25551 0.24412 &lt;q [132] &lt;0.24414 0.23300 &lt;q [133] &lt;0.23302 0.22211 &lt;q [134 ] &lt;0.22213 0.21143 &lt;q [135] &lt;0.21145 0.20095 &lt;q [136] &lt;0.20097 0.19067 &lt;q [137] &lt;0.19069 0.18058 &lt;q [138] &lt; 0.18060 0.17066 &lt;q [139] &lt;0.17068 0.16091 &lt;q [140] &lt;0.16093 0.15132 &lt;q [141] &lt;0.15134 0.14189 &lt;q [142] &lt;0, 14191 0.13259 &lt;q [143] &lt;0.13261 0.12344 &lt;q [144] &lt;0.12346 0.11442 &lt;q [145] &lt;0.11444

0.10552 &lt;q [146] &lt;0.10554 0.09675 &lt;q [147] &lt;0.09677 0.08809 &lt;q [148] &lt;0.08811 0.07954 &lt;q [149] &lt;0.07956 0, 07110 &lt;q [150] &lt;0.07112 0.06275 &lt;q [151] &lt;0.06277 0.05451 &lt;q [152] &lt;0.05453 0.04635 &lt;q [153] &lt;0.04637 0.03828 &lt; q [154] &lt;0.03830 0.03030 &lt;q [155] &lt;0.03032 0.02240 &lt;q [156] &lt;0.02242 0.01458 &lt;q [157] &lt;0.01460 0.00683 &lt;q [ 158] &lt;0.00685 -0.00084 &lt;q [159] &lt;-0.00082 -0.00845 &lt;q [160] &lt;-0.00843 -0.01598 &lt;q [161] &lt;-0.01596 -0 , 02345 &lt;q [162] &lt;-0.02343 -0.03085 &lt;q [163] &lt;-0.03083 -0.03819 &lt;q [164] &lt;-0.03817 -0.04547 &lt;q [165] &lt;-0.04545 -0.05269 &lt;q [166] &lt;-0.05267 -0.05984 &lt;q [167] &lt;-0.05982 -0.06694 &lt;q [168 ] &lt;-0.06692 -0.07397 &lt;q [169] &lt;-0.07395 -0.08095 &lt;q [170] &lt;- 0.08093 -0.08786 &lt;q [171] &lt;-0.08784 -0 , 09471 &lt;q [172] &lt;-0.09469 -0.10149 &lt;q [173] &lt;-0.10147 -0.10821 &lt;q [174] &lt;-0.10819 -0.11486 &lt;q [175] &lt; -0.1484 -0.12144 &lt;q [176] &lt;-0.12142 -0.12794 &lt;q [177] &lt;-0.12792 -0.13437 &lt;q [178] &lt;-0.13435 -0.14071 &lt;q [179] &lt;-0.14069 -0.114697 &lt;q [180] &lt;-0.14695 -0.15313 &lt;q [181] &lt;-0.15311 -0.15919 &lt;q [182] &lt; -0.15917 -0.16514 &lt;q [183] ​​&lt;-0.16512 -0.17097 &lt;q [184] &lt;-0.17095

-0.17668 &lt;q [185] &lt;-0.17666 -0.118226 &lt;q [186] &lt;-0.18224 -0.18768 &lt;q [187] &lt;-0.18766 -0.119295 &lt;q [188 ] &lt;-0.19293

-0.19804 &lt;q [189] &lt;-0.19802

-0.20294 &lt;q [190] &lt;-0.20292 -0.20764 &lt;q [191] &lt;-0.20762

Even more accurately, the filter coefficients q (y) can be expressed by the following equations io for integers in the range between 0 and 191, where, according to the requirements and specifications of the particular implementations, the prototype filter coefficients may differ from the following equations, either individually or from the maximum absolute value, typically by 10%, 5% or 2% and preferably by 1% or 0.1%:

i5 q [0] = -0.2029343380 q [l] = -0.1980331588 q [2] = -0.1929411519 q [3] = -0.1876744222 q [4] = -0.1822474011 q [5] = -0.1766730202 q [6] = -0.1709628636 q [7] = -0.1651273005 q [8] = -0.1591756024 q [9] = -0.1531160455 q [10] = -0.1469560005 q [ll] = -0.1407020132 q [12] = -0.1343598738 q [13] = -0.1279346790 q [14] = -0.1214308876 so q [15] = -0.1148523686 q [16] = -0.1082024454 q [17] = -0.1014839341 q [18] = -0.0946991783 q [19] = -0.0878500799 q [20] = -0.0809381268 q [21] = -0.0739644174 q [22] = -0.0669296831 q [23] = -0.0598343081 q [24] = -0.0526783466 q [25] = -0.0454615388 q [26] = -0.0381833249 q [27] = -0.0308428572 q [28] = -0.0234390115 q [29] = -0.0159703957 q [30] = -0.0084353584 q [31] = -0.0008319956 q [32] = 0.0068418435 q [33] = 0, 0145885527 io q [34] = 0.0224107648 q [35] = 0.0303113495 q [36] = 0.0382934126 q [37] = 0.0463602959 q [38] = 0.0545155789 is q [39] = 0, 0627630810 q [40] = 0.0711068657 q [41] = 0.0795512453 q [42] = 0.0881007879 q [43] = 0.0967603259 q [44] = 0.1055349658 q [45] = 0.1144301000 q [46] = 0.1234514222 q [47] = 0.1326049434 q [48] = 0.1418970123 q [49] = 0.1513343370 q [50] = 0.1609240126 q [51] = 0, 1706735517 q [52] = 0.1805909194 q [53] = 0.1906845753 so q [54] = 0.2009635191 q [55] = 0.2114373458 q [56] = 0.2221163080 q [57] = 0.2330113868 q [58] = 0.2441343742 q [59] = 0.2554979664 q [60] = 0.2671158700 q [61] = 0.2790029236 q [62] = 0.2911752349 q [63] = 0.3036503350 q [ 64] = 0.9025275713 q [65] = 0.9103585196 q [66] = 0.9176977825 q [67] = 0.9245760683 q [68] = 0.9310214581 q [69] = 0.9370596739 q [70] = 0.9427143143 q [71] = 0.9480070606 q [72] = 0.9529578566 io q [73] = 0.9575850672 q [74] = 0.9619056158 q [75} = 0.9659351065 q [76] = 0.9696879297 q [77] = 0.9731773547 is q [78] = 0.9764156119 q [79] = 0.9794139640 q [80] = 0.9821827692 q [81] = 0 , 9847315377 q [82] = 0.9870689790 q [83] = 0.9892030462 q [84] = 0.9911409728 q [85] = 0.9928893067 q [86] = 0.9944539395 q [87] = 0, 9958401318 q [88] = 0.9970525352 q [89] = 0.9980952118 q [90] = 0.9989716504 q [91] = 0.9996847806 q [92] = 1,0002369837 so q [93] = 1,0006301028 q [94] = 1,0008654482 q [95] = 1,0009438063 q [96] = 1,0008654482 q [97] = 1,0006301028 35 q [98] = 1,0002369837 q [99] = 0.9996847806 q [100] = 0.9989716504 qtlOl] = 0.9980952118 q [102] = 0.9970525352 q [103] = 0.9958401318 q [104] = 0.9944539395 q [105] = 0.9928893067 q [106] = 0.9911409728 q [107] = 0.9892030462 q [108] = 0.9870689790 q [109] = 0.9847315377 qtl 10] = 0.9821827692 q [l 11 ] = 0.9794139640 io q [l 12] = 0.9764156119 q [l 13] = 0.9731773547 q [l 14] = 0.9696879297 q [l 15] = 0.9659351065 q [l 16] = 0.9619056158 is q [l 17] = 0.9575850672 q [l 18] = 0.9529578566 q [119] = 0.9480070606 q [120] = 0.9427143143 q [121] = 0.9370596739 q [122] = 0.9310214581 q [123] = 0.9245760683 q [124] = 0.9176977825 q [125] = 0.9103585196 q [126] = 0 , 9025275713 25 q [127] = 0.8941712974 q [128] = 0.2911752349 q [129] = 0.2790029236 q [130] = 0.2671158700 q [131] = 0.2554979664 so q [132] = 0 , 2441343742 q [133] = 0.2330113868 q [134] = 0.2221163080 q [135] = 0.2114373458 q [136] = 0.2969635191 q [137] = 0.1906845753 q [138] = 0.1805909194 q [139] = 0.1706735517 q [140] = 0.1609240126 q [141] = 0.1513343370 q [142] = 0.1418970123 q [143] = 0 , 1326049434 q [144] = 0.1234514222 q [145] = 0.1144301000 q [146] = 0.1055349658 q [147] = 0.0967603259 q [148] = 0.0881007879 q [149] = 0.0795512453 q [150] = 0.0711068657 q [151] = 0.0627630810 q [152] = 0.0545155789 q [153] = 0.0463602959 q [154] = 0.0382934126 q [155] = 0.0303113495 is q [156] = 0.0224107648 q [157] = 0.0145885527 q [158] = 0.0068418435 q [159] = -0.0008319956 q [160] = -0.0084353584 q [161] = -0.0159703957 q [162] = -0.0234390115 q [163] = -0.0308428572 q [164] = -0.0381833249 q [165] = -0.0454615388 q [166] = -0.0526783466 q [167] = -0.0598343081 q [168] = -0.0669296831 q [169] = -0.0739644174 q [170] = -0.0809381268 so q [171] = -0.0878500799 q [172] = -0.0946991783 q [173] = -0.1014839341 q [174] = -0.1082024454 q [175] = -0.1148523686 q [176] = -0.1214308876 q [177] = -0.1279346790 q [178] = -0.1343598738 q [179] = -0.1407020132 q [180] = -0.1469560005 q [181] = -0.1531160455 q [182] = -0.1591756024 q [183] ​​= -0.1651273005 q [184] = -0.1709628636 q [185] = -0.1766730202 q [186] = -0.1822474011 q [187] = -0.1876744222 q [188] = -0.1929411519 q [189] = -0.1980331588 q [190] = -0.2029343380 q [191] = -0.2076267137

Accordingly, the invention relates to the application of any filter to a signal contained in the transformation domain of a complex exponentially modulated filter bank when this filter bank is designed to provide a virtual alias-free performance of operations, such as equation, spectral envelope alignment, frequency selective panning, or frequency selective spatialization of audio signals. The invention enables an efficient transformation of a given final pulse response (FIR) filter in the time domain to a set of shorter FIR filters used with a filter for each subband in the filter bank.

The invention describes how a given discrete time domain filter is converted to a set of subband domain filters. The result is that a given filter can be implemented with a high degree of accuracy in the subband domain of a complex, exponentially modulated filter bank. In a preferred embodiment, the filter converter consists of a second, complex exponentially modulated analysis filter bank. For filters which implement a clean delay, the methods of the invention coincide with PCT / EP2004 / 004607 Advanced processing based on a complex-exponential modulated filter bank and adaptive time framing.

Furthermore, the invention comprises the following features:

A method of obtaining a high quality approximation of the filtering of a discrete time input signal with a given filter comprising the steps of:

analyzing the input signal with a down-sampled, complex analysis filter bank to obtain multiple subband signals, filtering each subband signal with a subband filter, where the multiple subband filters are obtained from the given filter by a filter converter, synthesizing an output signal from the filtered subband signals,

A method according to the above wherein the filter converter consists of a downsampled, complex analysis filter bank.

An apparatus for performing the method of obtaining high quality approximation of the filtering of a discrete time input signal with a given filter comprising the steps of:

analyzing the input signal with a downsampled, complex analysis filter bank to obtain multiple subband signals, filtering each subband signal with a subband filter, the multiple subband filters being obtained from the given filter by a filter converter, synthesizing the output signal from the filtered subband signals,

A computer program with instructions for performing, when running on a computer, a method for achieving high quality approximation of the filtering of a discrete time input signal with a given filter, the method comprising the steps of:

analyzing the input signal with a down-sampled, complex analysis filter bank to obtain multiple subband signals, filtering each subband signal with a subband filter, where the multiple subband filters are obtained from the given filter by a filter converter, synthesizing an output signal from the filtered subband signals,

Adaptation for real cosine modulated filter banks

While the above derivation is based on complex, modulated filter banks, a remark should be made here for the critically sampled real rendering obtained by a cosine modulated filter bank as defined by taking the real portion of the subband samples (1) for an appropriate phase factor θ. In this case, it is no longer appropriate to use the in-band subband filtering method (3) to obtain a good approximation to a given filter. However, due to the assumptions made about the prototype filter response, a generalization of a multiband filter of type (22) r = -ll will be relevant (with obvious modifications for the first and last subbands). Due to the critical sampling, there will be much less freedom during the construction of filter machine g^rn (l). It is necessary to perform the following which will appear to a person skilled in the art. For each m = 0, l, ..., ll, the elementary subband signal d is used_n(k) = d [nm] d [k] as the signal to the real synthesis bank and filter the resulting output signal y (v) with the filter h (y) to obtain the filtered synthesis waveform z (y). This filtered waveform is sent to the real analysis bank. The resulting subband signal carries the coefficient of the masks g^rn (T) for n + r = m. Some reduction of the work required for the filter is achieved by observing that three cases m = 3 / c + a for ε = 0.1,2 can be treated in parallel by feeding the first synthesis bank with all the corresponding elementary subband signals for each case. Thus, the real valued filter converter includes these three real syntheses and three real analysis bank operations. This parallel calculation represents an implementation that is a truncation for truly valued filter converter for the case of QMF bands with good lateral disassembly.

Fig. 9 shows an embodiment of the new filter apparatus for filtering a time domain input signal of a new filter apparatus to obtain a time domain output signal. As already mentioned in connection with FIG. 1a, the filter apparatus of FIG. 9 a complex analysis filter bank 101, a subband filtering 102 and a complex synthesis filter bank 103 which transmits the time domain's output signal.

While FIG. 1 shows a system comprising an embodiment of the new filter apparatus together with an embodiment of a filter converter 104, the filter apparatus of FIG. 9 only as an option, a filter converter 104 which provides subband filtering 102 with intermediate filter definition signal e.g. in the form of filter outlets or pulse response for each of the intermediate filters 190 in the subband filtering 102. The filter apparatus of FIG. 9 includes any additional components capable of providing subband filtering 102 with filter outlets for the multiple intermediate filters 190 of subband filtering 102.

As an example, the filter outlets may also be taken from any database 500 connected to the subband filtering 102.1. An embodiment comprises the database 500 complex, valued filter outputs of the intermediate filters 190. The database can be implemented as a memory, e.g. in the form of a non-volatile memory or a volatile memory, depending on the specific implementation. Accordingly, memory solutions for database 500 may include ROM (ROM = read-only memory), RAM (RAM = arbitrary access memory), flash memory, magnetic memory, optical memory, or other memory system.

Depending on the specific implementation, a processor or CPU (CPU = central processing unit) 510 may access the database and provide filter outputs to the subband filtering 102 or may also access the database to provide corresponding filter outputs to the intermediate filters of the subband filtering 102. Accordingly, such an embodiment includes database 500 from which the filter outlets for the subband filtering can be taken.

In another embodiment of the new filter apparatus which is also shown as a choice in FIG. 9, CPU 510 can directly calculate the filter outlets. In such an embodiment, CPU 510 accesses database 500 according to a set of parameters provided by the user and / or according to a set of parameters based on other circumstances and reads one or more sets of filter outputs for the intermediate filters of the subband filtering 102 and calculates, optionally together with an interpolation system or other calculation system, the desired intermediate filter outlets or deliver these to the subband filtering 102. In another embodiment, CPU 510, or another processor or computer system, supplies the filter outlets for intermediate filters 190 to subband filtering 102 without accessing a database 500. In such an embodiment, CPU 510 or another processor calculates the filter outlets and supplies them to subband filtering 102. Such an embodiment will be explained in connection with FIG. 10.

In another embodiment as shown in FIG. 9, CPU 510 accesses another database 520 and reads one or more filter definition signals (e.g., in the form of pulse response signals corresponding to the filter characteristic in the time domain), calculates an effective filter definition signal, e.g. a suitable pulse response and delivers the results of this calculation to the filter converter 104. In this embodiment, the filter converter 104 then provides the subband filtering 102 with appropriate filter outlets for the intermediate filters 190. Accordingly, in this embodiment, filter converter 104 generates the effective subband filters or intermediate filters applied to each of the individual subband filters of each subband signal within the subband filtering 102 which results in a filtering effect that is audibly indistinguishable from the corresponding filter used in the time domain input signal. As a result, this embodiment is also capable of directly calculating the filter outlets via the filter converter 104.

An example may be e.g. be a device which calculates the outlets of the intermediate filters 190 of the subband filtering 102 in accordance with a set of parameters, e.g. as provided by the user, where the parameter basis is so large that an effective predetermination of the filter outlets, possibly together with a type of interpolation system, does not produce the desired results.

A more concrete application, e.g. be on the site with a dynamic chance for HRTF filters in a domain to be converted to the subband or QMF domain. As mentioned earlier, this is e.g. relevant in applications involving head tracking where database 520 is an HRTF database comprising time pulse responses of the HRTF filters. Since the HRTF filters usually have a very long pulse response, the use of such a system is particularly interesting as the outlets for the intermediate filters 190 or the QMF outputs are complex. Saving the database in this domain will roughly double the memory requirements compared to the memory requirement for storing the pulse responses in the time domain. However, the advantage of reduced memory requirements can also be used without a CPU 510 which calculates the pulse response delivered to filter converter 504.1 instead, database 520 may simply be prompted to send the corresponding definition signal, which may be a pulse response in the time domain, to filter converter 104.

In FIG. 20, an amplitude / frequency characteristic 550 is shown in the frequency domain. In some applications as previously explained, the filter coefficient or filter outlets are intermediate filters 190 of the subband filtering 102 and can be stored in the database similar to the database 500 of FIG. 9. Alternatively or additionally, the filter outlets of the intermediate filters in some applications may also be calculated by CPU 510 of FIG. 9. In the case of a special power filtering or a lower quality signal processing where foreign effects can be tolerated (at least to some extent), the filter outlets of the intermediate filters 190 after subband filtering 102 can be calculated without a filter converter 104 or another embodiment of a filter converter. Possible uses include, in particular, voice transmission over low-quality wires, such as telephones or small band radio communications. Accordingly, in such applications, a filter outlet determination corresponding to the transfer function 550 of FIG. 10 or another amplitude / frequency characteristic of several sub-bands 560 with different sub-band frequencies is performed without using the inventive filter converter.

Fig. 11 shows a filter converter 104 according to the invention. As previously outlined in connection with FIG. 3, the filter converter 104 comprises a complex analysis filter bank 301 to which a (truly valued) pulse response signal indicating an amplitude / frequency filter characteristic can be delivered via an input 104a and via an optional switch 600. As outlined earlier, the complex analysis filter bank 301 converts the pulse response signal into multiple subband signals with complex values ​​and the mehom filter definition signal at an output 104b of the filter converter. As shown in FIG. 1a and fig. 9, the output 104b of the filter converter 104 can be connected to a subband filtering 102.

As already mentioned, each of the complex valued subband signals at the complex modulated filter bank 301 corresponds to a pulse response for one of the mehom filters 190 for a subband signal in the subband filtering 102 shown in FIG. 1a and 9. Typically, the complex valued subband signals are significantly shorter than the pulse response signal of the filter characteristic delivered at input 109a in the time domain. Furthermore, typically at least one of the complex valued subband signals at output 104 comprises at least two different non-vanishing values. In particular, the last feature signal from filter converter 104 separates out a single gain adjustment in the filtering framework using a direct Fourier transform procedure.

However, if filter converter 104 is not provided with a pulse response signal indicating an amplitude / frequency filter characteristic, but a filter definition signal comprising at least either an amplitude / frequency filter characteristic, a phase / frequency filter characteristic of the filter outputs in the time domain, or another domain of a filter, for example, the filter converter 104 comprises a pulse response generator 610 for converting the filter definition signal to the pulse response signal which is then delivered via the optional switch 600 to the complex analysis filter bank 301.1. the pulse response generator 610 e.g. calculating the pulse response signal delivered to complex analysis filter bank 301 by overlaying truly valued oscillations (Fourier synthesis), where the amplitude characteristics and phase characteristics of the intended filter transmitted to the complex subband domain are considered defined by the definition signal supplied to input 104c. If at least either an amplitude / frequency characteristic or a phase / frequency characteristic is applied to the pulse response generator

610, in other words, a pulse response signal can be calculated by pulse response generator 610 by superimposing (harmonic) oscillations taking into account the amplitude and phase relationships as defined by the filter definition signal.

Possible uses of both designs of the filter apparatus and the filter converter and especially in the field of quality audio decoding and decoding

Recent developments in audio coding have provided a device for obtaining a multi-channel signal impression over stereo headphones. This is usually done by mixing a multichannel signal to stereo by braking the original multichannel signal and HRTF filters. It has been found in the present art that the multi-channel parametric audio decoder can be combined with a two-channel downmix algorithm which allows to output a multi-channel signal and headphones without the need to first reproduce the multi-channel signal from the transmitted downmixed signal and then re-mix it using of HRTF filters. However, this requires that the parameters to recreate the multichannel signal (e.g. IID, CLD parameters) are combined with HRTF filters which in turn require parameterization of HRTF filters. This requirement for parameterization of HRTF filters causes a major limitation in the system since HRTF filters can be long and thus very difficult to properly modulate with a parametric method. This limitation makes it impossible to brake long HRTF filters for combined, parametric multi-channel and two-channel downmix decoders. The important algorithmic component required to obtain a proper combination of multi-channel parameters and HRTF filters is to access a reproduction of the given HRTF filters in the subband domain with the spatial parameters. This is exactly what is offered by the embodiment of the invention. After this rendering becomes possible, the HRTF filters can be combined into 2 / V filters as a function of the multi-channel parametric rendering. This provides a significant advantage of computational complexity compared to the method that first recreates the M channels and then breaks the A7 fi ltration.

An example of another application of the method brought about by the practice of the invention is the effective compensation for non-perfect audio reproducing devices for audio content encoded in MPEG HE-AAC format [ISO / IEC 14496-3: 2001 / AMDl: 2003]. Such advanced filtration steps, which may include cross-talk cancellation, can be used directly in the subband domain prior to time domain synthesis.

Other developments in audio coding have led to methods for recreating multichannel reproduction of an audio signal based on a stereo (or mono) signal and corresponding control data. These methods differ substantially from an older matrix-based solution, such as e.g. Dolby Prologic, since additional control data is transmitted to control the reproduction and is also called mixing the surround channels based on the transmitted mono or stereo channels.

Accordingly, such a parametric multichannel audio decoder, e.g. MPEGsurround N channels based on M transmitted channels, where M> N and additional control data. This additional control data represents a significantly lower data rate than is required for transmitting all / V channels and makes coding very efficient while ensuring compatibility with both M-channel devices and / V-channel devic es. [J. Breebaart et al. MPEG spatial audio coding / MPEG Surround: overview and current status, Proc. 119th AES Convention, New York, USA, October 2005, Preprint 6447],

These parametric surround coding methods typically include a surround signal parameterization based on channel level difference (CLD) and interchannel context / cross correlation (ICC). These parameters describe power ratios and correlation between the channel pairs in the mixing process. Furthermore, channel prediction coefficients (CPCs) are also used by the current technique to predict or predict intermediate or output channels during the upbeat procedure.

Depending on the particular implementation of the methods according to the invention, the new methods may be implemented in hardware or in software. The implementation can be carried out using a digital storage medium or, in particular, a disk, CD or DVD with an electronically readable control signal that interacts with a programmable co mputer system by carrying out an execution of the new methods. Accordingly, in general, an embodiment of the invention is a computer program product having a program code stored on a machine-readable carrier, the program code being used to perform the new methods when running the program product on a computer or from a processor. In other words, the execution of the method according to the invention is accordingly a computer program with a program code for performing at least one of the new methods when the computer program is run on a computer.