MXPA00001837A - Reducing sparseness in coded speech signals - Google Patents

Abstract

Sparseness is reduced in an input digital signal (A) which includes a first sequence of sample values. An output digital signal (B) is produced in response to the input digital signal. The output digital signal includes a second sequence of sample values, which second sequence of sample values has a greater density of non-zero sample values than the first sequence of sample values.

Description

REDUCTION OF DISPERSION IN CODIFIED VOICE SIGNALS This application claims priority according to 35 USC 119 (e) (l) of the pending US Provisional Application, No. 06 / 057,752, filed on September 2, 1997, and is a partial continuation of the pending application Serial No. from the USA 09 / 0345,590 (file 34645-405), presented on March 4, 1998.

FIELD OF THE INVENTION The invention relates, generally, to voice coding and, more particularly, to the problem of dispersion in encoded speech signals.

BACKGROUND OF THE INVENTION Voice coding is an important part of modern digital communications systems, for example, wireless radio communication systems, such as cellular, digital telecommunications systems. To achieve the high capacity required by such systems, now and in the future, it is imperative to provide efficient compression of voice signals, while also supplying high quality voice signals. In this aspect, when the bit rate of a speech coder is decreased, for example, to provide an additional communications channel capacity for other communication signals, it is desirable to obtain an elegant degradation of speech quality, without introducing intrusive artifacts. . Conventional examples of lower rate voice coders for cellular telecommunications are illustrated in IS-641 (D-AMPS EFR) and by the normal G.729ITU. The encoders specified in the above standards are similar in structure, including both an algebraic codebook, which typically supplies a relatively sparse output. Dispersion refers, in general, to the situation where only a few of the samples of an entry in a given codebook have a non-zero sample value. This dispersion condition is particularly frequent when the bit rate of the algebraic codebook is reduced in an attempt to supply the compression of the speech. With very few non-zero samples in the codebook to initiate, and with the lower bit rate requiring that even fewer codebook samples be used, the resulting dispersion is an easily perceived degradation in the coded speech signals of the conventional voice coders, mentioned above. Therefore, it is convenient to avoid the aforementioned degradation in coded speech signals, when the bit rate of a speech coder is reduced to provide speech compression. In an attempt to avoid the aforementioned degradation in coded speech signals, the present invention provides an anti-spreading operator, to reduce the dispersion in an encoded speech signal, or any digital signal, where this dispersion is disadvantageous.

BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a block diagram illustrating an example of an anti-dispersion operator of the present invention; Figure 2 illustrates various positions in a Code Excited Linear Predictive encoder / decoder, where the anti-scatter operator of Figure 1 can be applied; Figure 2A illustrates a communications transceiver (transmitter-receiver), which can use the structure of the encoder / decoder of Figures 2 and 2B; Figure 2B illustrates another exemplary Code Excited Linear Predictive decoder, including the anti-scatter operator of Figure 1; Figure 3 illustrates an example of the anti-scattering operator of Figure 1; Figure 4 illustrates an example of how the additive signal of Figure 3 can be produced; Figure 5 illustrates in a block diagram form, how the anti-scattering operator of Figure 1 can be incorporated as an anti-scattering filter; Figure 6 illustrates an example of the anti-scattering filter of Figure 5; Figures 7 to 11 graphically illustrate the operation of an anti-dispersion filter of the type illustrated in Figure 6; Figures 12-16 graphically illustrate the operation of an anti-dispersion filter of the type illustrated in Figure 6 and a relatively lower level of the anti-dispersion operation than the anti-dispersion filter of Figures 7 to 11; Figure 17 illustrates another example of the anti-scattering operator of Figure 1; and Figure 18 illustrates an exemplary method of providing the anti-dispersion modification, according to the invention.

DETAILED DESCRIPTION Figure 1 illustrates an example of an anti-dispersion operator, according to the present invention. The anti-spreading operator ASO of Figure 1 receives at its input A a dispersion, a digital signal received from a source 11. The anti-spreading operator ASO operates on signal A and supplies a digital signal B at its output, which is less scattered than the input signal A. Figure 2 illustrates several example locations where the ASO anti-spreading operator of Figure 1 can be applied to a Code Excited Linear Predictive (CELP) speech encoder, provided on a transmitter, for use in a wireless communication system, or in a CELP speech decoder provided in a receiver of a wireless communication system. As shown in Figure 2, the anti-dispersion operator ASO can be provided at the output of the fixed codebook 21 (for example algebraic), and / or at any other location designated by the reference numbers 201-206. . In each of the locations designated in Figure 2, the anti-spreading operator ASO of Figure 1 will receive the spreading signal at its input A and will supply at its output B a smaller spreading signal. Thus, the structure of the CELP decoder / decoder, shown in Figure 2, includes several examples of the scattering signal source of Figure 1. The broken line in Figure 2 illustrates the conventional feedback path to the adapted codebook, as conventionally supplied in the CELP voice coders / decoders. If the anti-spreading operator ASO is provided where shown in Figure 2 and / or at any of the locations 201-204, then one or more anti-spreading operators will affect the coded excitation signal reconstructed by the decoder at the output of the 210 circuit of sum. If applied at locations 205 and / or 206, one or more anti-spreading operators will have no effect on the output of the coded excitation signal from the summing circuit 210. Figure 2B illustrates an exemplary CELP decoder, which includes an additional sum circuit 25, which receives the outputs of the code books 21 and 213, and supplies the feedback signal from the adapted codebook 23. If the operator anti- ASO dispersion is provided where shown in Figure 2B, and / or in locations 220 and 240, then such anti-dispersion operators will not effect the feedback signal to the adapted codebook 23. Figure 2A illustrates a transceiver, whose receiver (RCVR) includes a structure of the CELP decoder of Figure 2 (or Figure 2B), and whose transmitter (XMTR) includes the structure of the CELP coder of Figure 2. Figure 2A illustrates that the transmitter receives as an input an acoustic signal and supplies as output to the reconstruction information of the communication channel which receiver can reconstruct the acoustic signal. The receiver receives input from the reconstruction information of the communication channel, and supplies a reconstructed acoustic signal. The illustrated transceiver and communication channel may be, for example, a transceiver in a cellular telephone and the air interface of a cellular telephone network, respectively. Figure 3 illustrates an exemplary embodiment of the ASO anti-spreading operator of Figure 1. In Figure 3, a noise-type signal m (n) is added to the spreading signal as received in A. Figure 4 illustrates an example of how a signal m (n) can be produced. A noise signal with a Gaussian distribution N (0,1) is filtered by a suitable high-pass filter and spectral coloration, to produce the noise-type signal m (n). As illustrated in Figure 3, the signal m (n) can be applied to the summation circuit 31 with a suitable gain factor via the multiplier 33. This gain factor of Figure 3 can be a fixed gain factor, the gain factor of Figure 3 may also be a function of the gain applied conventionally to the output of the adapted codebook 23 (or a similar parameter describing the amount of the periodicity). In an example, the gain of Figure 3 will be 0 if the gain of the adapted codebook exceeds a predetermined threshold, and linearly increases as the gain of the codebook adapted from the threshold decreases. The gain of Figure 3 can also be realized analogously as a function of the gain applied conventionally to the output of the fixed codebook 21 of Figure 2. The gain of Figure 3 can also be based on a corresponding power spectrum. to the signal m (n) for the target signal used in the conventional voice method, in this case, the gain will need to be encoded and transmitted to the receiver. In another example, the addition of a noise type signal can be performed in the frequency domain in order to obtain the benefit of advanced frequency domain analysis. Figure 5 illustrates another exemplary embodiment of the ASO of Figure 2. The arrangement of Figure 5 can be characterized as an anti-dispersion filter designed to reduce the dispersion in the digital signal received from the source 11 of Figure 1. A example of the anti-scattering filter of Figure 5 is illustrated in greater detail in Figure 6. The anti-scattering filter of Figure 6 includes a winding section 53, which executes a coiling of the coded signal received from the codebook 21 fixed (for example algebraic), with an impulse response (at 65) associated with a full-pass filter. The operation of an example of the anti-scattering filter of Figure 6 is illustrated in Figures 7 to 11. Figure 10 illustrates an example of an entry from the codebook 21 of Figure 2, which has only two samples not of zero out of a total of forty samples. This dispersion characteristic will be reduced if the number (density) of non-zero samples can be increased. One way to increase the number of non-zero samples is to apply the codebook entry of Figure 10 to a filter having a suitable characteristic, to disperse the energy through the block of the forty samples. Figures 7 and 8 illustrate, respectively, the phase magnitude (in radians) characteristic of an all-pass filter, which can be operated to properly disperse the energy through the forty samples of the codebook entry of the Figure 10. The filter of Figures 7 and 8 alters the phase spectrum in the high frequency area between 2 and 4 kHz, while altering the low frequency areas below 2 kHz, only very marginally. The magnitude spectrum remains essentially unchanged by the filter of Figures 7 and 8. The example of Figure 9 graphically illustrates the impulse response of the all-pass filter, defined by Figures 7 and 8. The anti-scattering filter of FIG. Figure 6 produces an impulse response winding of Figure 9 in the block of Figure 10 of the samples. Because the codebook entries are provided from the codebook as blocks of forty samples, the rolling operation is performed in the form of blocks. Each sample in Figure 10 will produce 40 intermediate multiplication results in the winding operation. Taking the sample at position 7 in Figure 10 as an example, the first 34 multiplication results are signaled to positions 7 to 40 of the results block of Figure 11 and the remaining 6 multiplication results are "wrapped around" according to a circular winding operation, so that they are assigned to positions 1 to 6 of the resulting block. The intermediate multiplication results, produced by each of the remaining samples of Figure 10, are assigned to the positions in the block of the results of Figure 11 in an analogous manner, and sample 1, of course, does not need winding around it. For each position in the result block of Figure 11, the 40 intermediate multiplication results assigned there (one multiplication result per sample in Figure 10) are added together, and the sum represents the rolling result for that position. It is clear from the inspection of Figures 10 and 11 that the circular winding operation alters the Fourier spectrum of the block of Figure 10, so that the energy is dispersed through the blog, thus drastically increasing the number (or density) of non-zero samples in the block, and correspondingly reducing the amount of dispersion. The effects of performing the circular winding on a block-by-block basis can be uniformed by the synthesis filter 211 of Figure 2. Figures 12 to 16 illustrate another example of the operation of an anti-dispersion filter of the type generally shown in Figure 6. The all-pass filter of Figures 12 and 13 alters the phase spectrum between 3 and 4 kHz, without substantially altering the phase spectrum below 3 kHz. The response of the filter pulse is shown in Figure 14. With reference to the block of results of Figure 16 and noting that Figure 15 illustrates the same block of samples as Figure 10, it is clear that the anti-dispersion operation illustrated in FIG. Figures 12 to 16 do not disperse the energy as much as in Figure 11. Thus, Figures 12 to 16 define an anti-dispersion filter that modifies the entry of the codebook less than the filter defined by Figures 7 through 11. therefore, the filters of Figures 7 to 11 and Figures 12 to 16 respectively define different levels of anti-dispersion filtration. A low gain value of the adapted codebook indicates that the adapted codebook component of the reconstructed excitation signal (output of summing circuit 210) will be relatively small, thus giving rise to the possibility of a relatively large contribution from the fixed codes 21 (for example algebraic). Due to the aforementioned dispersion of the entries of the fixed codebook, it will be advantageous to select the anti-scatter filter of Figures 76 to 11, rather than those of Figures 12 to 16, because the filter of the Figures 7 to 11 provide a greater modification of the sample block than the filter of Figures 12 to 16. With values greater than the gain of the adapted codebook, the contribution of the fixed codebook is relatively smaller, so the filter of Figures 12 to 16, which provides less anti-dispersion modification, can be used. The present invention thus provides the ability to use the local characteristics of a given speech segment to determine if and how much to modify the characteristic dispersion associated with that segment. The windings made in the anti-scattering filter of Figure 6 can also be a linear winding, which provides a more uniform operation, because the effects of the block-type process are avoided. Also, although the batch process was described in the previous examples, such a process is not required in the practice of the invention, it is merely a feature of the conventional CELP decoder / decoder structure, shown in the examples. A closed-loop version of the method can be used. In this case, the encoder takes into account the anti-dispersion modification during the search of the code books. This will provide improved performance at the expense of increased complexity. The winding operation (circular or linear) can be performed by multiplying the filter matrix constructed from the conventional impulse response of the search filter by a matrix defining the anti-dispersion filter (using linear or circular winding). Figure 7 illustrates another example of the ASO anti-spreading operator of Figure 1. In the Example of Figure 17, an anti-spreading filter, of the type illustrated in Figure 5, receives the input signal A, and the output of the Anti-dispersion filter is multiplied by 170 by a gain factor g2. The noise-like signal m (n) of Figures 3 and 4 is multiplied by 172 by the gain factor g ^, and the outputs of the multipliers g1 and g, 170 and 172, are added together in 174 to produce the output signal B. The gain factors gi and g2 can be determined, for example, as follows. The gain gi can first be determined in one of the above described ways, with respect to the gain in Figure 3, and then the gain factor g2 can be determined as a function of the gain factor g ±. For example, the gain factor g2 can vary inversely with the gain factor g ^. Alternatively, the gain factor g can be determined in the same way as the gain of Figure 3, and then the gain factor g can be determined as a function of the gain factor g2, for example g ^ can vary inversely with g2. In an example of the arrangement of Figure 17, the anti-scattering filter of Figures 12-16 is used; the gain factor g2 = 1; m (n) is obtained by normalizing the distribution of Gaussian noise N (0,1) of Figure 4, to have an energy level equal to the entries of the fixed codebook, and adjusting the trimming frequency of the high pass filter of Figure 5 at 2000 Hz; and the gain factor gi is 80% of the gain of the fixed codebook. Figure 18 illustrates an exemplary method of providing the anti-dispersion modification, according to the invention. At 181, the dispersion level of the coded speech signal is estimated. This can be done offline or in an adapted form during the voice process. For example, algebraic codebooks and multi-pulse codebooks, the samples may be close to each other or separate, resulting in a variable dispersion; while in a regular pulse code book, the distance between the samples is fixed, so the dispersion is constant. In 183, an appropriate level of anti-dispersion modification is determined. This step can also be carried out offline or in an adapted form during the speech process, as described above. As another example of determining in an adapted form the anti-dispersion level, the impulse response (see Figures 6, 9 and 14) can be changed from one block to another. At 185, the selected level of anti-scatter modification is applied to the signal. It will be apparent to those skilled in the art, that the embodiments described above, with respect to Figures 1 to 18, can be carried out easily, using, for example, a properly programmed digital signal processor or other data processor. , and may alternatively be enhanced using, for example, each appropriately programmed digital signal processor or other data processor, in combination with the additional external circuitry there connected. Although exemplary embodiments of the invention have been described in detail above, they do not limit the scope of the invention, which can be practiced in a variety of embodiments.

Claims (28)

1. CLAIMS 1. An apparatus for reducing the dispersion in a digital input signal, which includes a first sequence of sample values, this apparatus comprises: an input for receiving the digital input signal; an anti-dispersion operator, coupled to the input and sensitive to the digital input signal, to produce a digital output signal, which includes a further sequence of sample values, this additional sequence of the sample values has a higher density of non-zero sample values, compared to the first sample values; and an output, coupled to the anti-dispersion operator, to receive the digital output signal therefrom. The apparatus of claim 1, wherein the anti-dispersion operator includes a circuit for adding a noise-like signal to the digital input signal. 3. The apparatus of claim 1, wherein the anti-dispersion operator includes a filter coupled to the input to filter the digital input signal. 4. The apparatus of claim 3, wherein the filter is a full-pass filter. 5. The apparatus of claim 3, wherein the filter uses circular winding and linear winding, to filter respective blocks of sample values in the first sequence of sample values. 6. The apparatus of claim 3, wherein the filter modifies a phase spectrum of the digital input signal, but leaves its spectrum of magnitude substantially unaltered. 7. The apparatus of claim 1, wherein the anti-distribution operator includes a signal path, extending from the input to the output, this signal path includes a filter, and the anti-spreading operator also includes a circuit to add a noise-like signal to a signal carried by the signal path. 8. The apparatus of claim 7, wherein the filter is a full-pass filter. 9. The apparatus of claim 7, wherein the filter uses one of the circular winding and the linear winding, to filter respective blocks of the sample values in the first sequence of these sample values. 10. The apparatus of claim 7, wherein the filter modifies a phase spectrum of the digital input signal, but leaves its spectrum of magnitude substantially unaltered. 11. An apparatus for the acoustic signal information process, this apparatus comprises: an input to receive the acoustic signal information; an encoder apparatus, coupled to the input and responsive to the information, to supply a digital signal, this digital signal includes a first sequence of the sample values; and an anti-dispersion operator, having an input coupled to the coding apparatus and responsive to the digital signal, to produce a digital output signal, which includes a second sequence of sample values, this second sequence of sample values has a greater density of non-zero sample values, compared to the first sequence of sample values. 12- The apparatus of claim 11. in which the encoding apparatus includes a plurality of codebooks, a summing circuit and a synthesis filter, these codebooks have respective outputs coupled to the respective inputs of the summing circuit, and this summing circuit has an output coupled to an input of the synthesis filter. 13. The apparatus of claim 12, wherein the anti-dispersion operator input is coupled to one of the outputs of the codebook. 14. The apparatus of claim 12, wherein the input of the anti-dispersion operator is coupled to the output of the summing circuit. 15. The apparatus of claim 12, wherein the anti-dispersion operator input is coupled to an output of the synthesis filter. 16. The apparatus of claim 12, wherein the coding apparatus is an encoding apparatus and the acoustic signal information includes an acoustic signal. 17. The apparatus of claim 12, wherein the coding apparatus is an encoding apparatus and the acoustic signal information includes the information from which an acoustic signal is to be constructed. 18. A method for reducing dispersion in a digital input signal, which includes a first sequence of sample values, this method comprises: receiving the digital input signal; produce, in response to the digital input signal., a digital output signal, which includes a second sequence of sample values, this second sequence of sample values has a higher density of non-zero sample values, in comparison with the first sequence of sample values; and produce the digital output signal. 19. The method of claim 18, wherein the production step includes filtering the digital input signal. 20. The method of claim 19. wherein the filtering step includes using a full-pass filter. 21. The method of claim 19, wherein the filtering step includes using a circular winding and a linear winding, to filter respective blocks of the sample values in the first sequence of sample values. 22. The method of claim 19, wherein the filtering step includes modifying a phase spectrum of the digital input signal, but leaving its magnitude spectrum substantially undisturbed. The method of claim 18, wherein the production step includes filtering a first signal to obtain a filtered signal, and adding a noise-like signal to one of the first signal and the filtered signal. 24. The method of claim 23, wherein the filtering step includes using a full-pass filter. 25. The method of claim 23, wherein the filtering step includes using a circular winding and a linear winding to filter the respective blocks of the sample values in the first sequence of the sample values. 26. The method of claim 23, wherein the filtering step includes modifying a phase spectrum of the digital input signal, but leaving its spectrum of magnitude substantially undisturbed. 27. The method of claim 18, wherein the production step includes adding a noise-type signal to the digital input signal. 28. A method for processing the acoustic signal information, this method comprises: receive the information of the acoustic signal; supplying, in response to the information, a digital signal, which includes a first sequence of the sample values; and produce, in response to the digital signal, a digital output signal, which includes one more sequence of the sample values, this additional sequence of the sample values has a higher density of the non-zero sample values, in comparison with the first sequence of the sample values.