JP4660496B2 - Speech coding apparatus and speech coding method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To prevent quality deterioration when a pitch period and a repetition period are different from each other. <P>SOLUTION: A period preselecting means 23 multiplies a repetition period of an adaptive sound source by a plurality of constants to find repetition period candidates of a plurality of driving sound sources, and selects repetition period candidates of a predetermined number of driving sound sources. A driving sound source encoding means 27 outputs the polarity of a sound source position which minimizes encoding distortion and an evaluated value of the encoding distortion at the time by the repetition period candidates of every predetermined number of driving sound sources. A period encoding means 28 compares evaluated values of encoding distortion by repetition periods, selects repetition period candidates of the driving sound sources based upon the comparison results, and outputs selection information, a sound source position code, and a polarity. <P>COPYRIGHT: (C)2007,JPO&amp;INPIT

Description

  The present invention relates to a speech encoding apparatus and speech encoding method for compressing a digital speech signal to a small amount of information.

  In many conventional speech coding apparatuses and speech decoding apparatuses, input speech is divided into spectral envelope information and a sound source, and each speech is encoded in units of frames of a predetermined long section, and this speech code is decoded. And the decoded speech is obtained by combining the spectral envelope information and the sound source by the synthesis filter. As the most typical speech coding apparatus and speech decoding apparatus, there is one using a code-driven linear prediction encoding (Code-Linear Prediction: CELP) system.

FIG. 14 is a block diagram showing the configuration of a conventional CELP speech encoding apparatus, and FIG. 15 is a block diagram showing the configuration of a conventional CELP speech decoding apparatus.
14 and 15, 1 is input speech, 2 is linear prediction analysis means, 3 is linear prediction coefficient coding means, 4 is adaptive excitation coding means, 5 is driving excitation coding means, and 6 is gain coding means. , 7 is a multiplexing means, 8 is a speech code, 9 is a separation means, 10 is a linear prediction coefficient decoding means, 11 is an adaptive excitation decoding means, 12 is a driving excitation decoding means, 13 is a gain decoding means, 14 Is a synthesis filter, and 15 is an output voice.

Next, the operation will be described.
In this conventional speech encoding device and speech decoding device, processing is performed in units of frames, with about 5 to 50 ms as one frame. First, in the speech coding apparatus shown in FIG. 14, input speech 1 is input to linear prediction analysis means 2, adaptive excitation coding means 4, and gain coding means 6. The linear prediction analysis means 2 analyzes the input speech 1 and extracts linear prediction coefficients that are speech spectral envelope information. The linear prediction coefficient encoding unit 3 encodes the linear prediction coefficient, outputs the code to the multiplexing unit 7, and outputs a linear prediction coefficient quantized for encoding the sound source.

  The adaptive excitation coding means 4 stores a sound source (signal) of a predetermined length in the past as an adaptive excitation codebook, and corresponds to each adaptive excitation code indicated by a binary value of several bits generated internally. Then, a time series vector in which past sound sources are periodically repeated is generated. Next, each time series vector is multiplied by an appropriate gain and passed through a synthesis filter using the quantized linear prediction coefficient output from the linear prediction coefficient encoding means 3 to obtain a temporary synthesized sound. The distance between the temporary synthesized sound and the input speech 1 is checked, an adaptive excitation code that minimizes this distance is selected and output to the multiplexing means 7, and a time series vector corresponding to the selected adaptive excitation code is obtained. The adaptive excitation is output to the drive excitation encoding means 5 and the gain encoding means 6. Further, a signal obtained by subtracting the synthesized sound from the adaptive sound source from the input sound 1 or the input sound 1 is output to the driving sound source encoding means 5 as a signal to be encoded.

First, the driving excitation encoding means 5 sequentially reads out the time series vectors from the driving excitation codebook stored therein corresponding to each driving excitation code indicated by a binary value of several bits generated internally. . Next, each of the read time series vectors and the adaptive excitation output from the adaptive excitation encoding unit 4 are multiplied by an appropriate gain and added, and the quantized linear prediction coefficient output from the linear prediction coefficient encoding unit 3 is added. A tentative synthesized sound is obtained by passing through a synthesis filter using. The distance between the provisional synthesized sound and the input signal 1 output from the adaptive excitation coding means 4 or the signal to be encoded which is a signal obtained by subtracting the synthesized sound from the adaptive sound source from the input sound 1 is checked, and this distance is minimized. Is selected and output to the multiplexing means 7, and a time series vector corresponding to the selected driving excitation code is output to the gain encoding means 6 as a driving excitation.

  First, the gain encoding means 6 sequentially reads out the gain vector from the gain codebook stored therein corresponding to each gain code indicated by a binary value of several bits generated internally. Then, each element of each gain vector is multiplied by the adaptive excitation output from the adaptive excitation encoding unit 4 and the driving excitation output from the driving excitation encoding unit 5 and added to generate a excitation, and the generated excitation is By passing through a synthesis filter using the quantized linear prediction coefficient output from the linear prediction coefficient encoding means 3, a provisional synthesized sound is obtained. The distance between the provisional synthesized sound and the input speech 1 is checked, and a gain code that minimizes this distance is selected and output to the multiplexing means 7. Further, the generated excitation corresponding to the gain code is output to the adaptive excitation encoding means 4.

  Finally, the adaptive excitation encoding unit 4 updates the internal adaptive excitation codebook using the excitation corresponding to the gain code generated by the gain encoding unit 6.

  The multiplexing unit 7 includes the code of the linear prediction coefficient output from the linear prediction coefficient encoding unit 3, the adaptive excitation code output from the adaptive excitation encoding unit 4, and the driving output from the driving excitation encoding unit 5. The excitation code and the gain code output from the gain encoding means 6 are multiplexed, and the obtained speech code 8 is output.

  Next, in the speech decoding apparatus shown in FIG. 15, the separating means 9 separates the speech code 8 output from the speech encoding apparatus and outputs the code of the linear prediction coefficient to the linear prediction coefficient decoding means 10. The adaptive excitation code is output to the adaptive excitation decoding unit 11, the driving excitation code is output to the driving excitation decoding unit 12, and the gain code is output to the gain decoding unit 13. The linear prediction coefficient decoding means 10 decodes the linear prediction coefficient from the code of the linear prediction coefficient separated by the separation means 9, sets it as a filter coefficient of the synthesis filter 14, and outputs it.

  Next, the adaptive excitation decoding means 11 stores the past excitation as an adaptive excitation codebook and periodically repeats the past excitation corresponding to the adaptive excitation code separated by the separating means 9. A sequence vector is output as an adaptive sound source. Further, the driving excitation decoding unit 12 outputs a time series vector corresponding to the driving excitation code separated by the separating unit 9 as a driving excitation. The gain decoding unit 13 outputs a gain vector corresponding to the gain code separated by the separating unit 9. Then, a sound source is generated by multiplying the two time-series vectors by each element of the gain vector and adding them, and an output sound 15 is generated by passing the sound source through the synthesis filter 14. Finally, adaptive excitation decoding means 11 updates the internal adaptive excitation codebook using the generated excitation.

Next, conventional techniques for improving the CELP speech coding apparatus and speech decoding apparatus will be described.
Non-Patent Document 1 discloses a CELP speech encoding apparatus and speech decoding apparatus in which a pulse excitation is introduced into the encoding of a driving excitation for the main purpose of reducing the amount of calculation and the amount of memory. In this conventional configuration, the driving sound source is expressed only by position information and polarity information of several pulses. Such a sound source is called an algebraic sound source and has a good coding characteristic for its simple structure, and has been adopted in many recent standard systems.

  FIG. 16 is a table showing pulse excitation position candidates used in Non-Patent Document 1. In the speech encoding apparatus shown in FIG. 14, the driving excitation encoding apparatus 5 is used, and in the speech decoding apparatus shown in FIG. It is mounted on the drive excitation decoding device 12. In Non-Patent Document 1, the excitation encoding frame length is 40 samples, and the driving excitation is composed of four pulses. As shown in FIG. 16, the position candidates of the sound source numbers 1 to 3 are restricted to 8 positions, and each pulse position can be encoded with 3 bits. The pulse of the sound source number 4 is restricted to the position of 16, and the pulse position can be encoded with 4 bits. By constraining the pulse sound source position candidates, it is possible to reduce the amount of calculation by reducing the number of encoded bits and the number of combinations while suppressing deterioration of the encoding characteristics.

  In Non-Patent Document 1, in order to reduce the calculation amount of pulse position search, the impulse response (synthesized sound by a single pulse sound source), the correlation function of the signal to be encoded, and the impulse response (by a single pulse sound source) The cross-correlation function of the synthesized sound) is calculated in advance and stored as a pre-table, and distance (encoding distortion) calculation is performed by simple addition of these values. Then, the pulse position and polarity that minimize this distance are searched. This process is performed by the driving excitation encoding device 5 of the speech encoding device shown in FIG.

The search method used in Non-Patent Document 1 will be specifically described below.
First, the distance minimization is equivalent to maximizing the evaluation value D expressed by the following equation (1), and the search is performed by calculating this evaluation value D for all combinations of pulse positions. Can be executed.
D = C ² / E (1)
However,

here,
m _k is the pulse position of the k-th pulse,
g (k) is the pulse amplitude of the kth pulse,
d (x) is the correlation value between the impulse response when the impulse is set at the pulse position x and the signal to be encoded,
φ (x, y) is a correlation value between an impulse response when an impulse is set at the pulse position x and an impulse response when an impulse is set at the pulse position y.

Further, in Non-Patent Document 1, g (k) is the same sign as d (m _k ) and the absolute value is 1, and the above equations (2) and (3) The calculation is simplified as shown in the equation.

However,
d ′ (m _k ) = | d (m _k ) | (6)
φ ′ (m _k , m _i )
= Sign [d (m _k )] sign [d (m _i )] φ (m _k , m _i ) (7)
If the calculation of d ′ and φ ′ is performed before the calculation of the evaluation value D for all combinations of pulse positions, the subsequent calculation is performed with a small amount of calculation such as simple addition of the equations (4) and (5). An evaluation value D can be calculated.

  A configuration for improving the quality of the algebraic sound source is disclosed in Patent Document 1 and Patent Document 2, and also disclosed in Non-Patent Document 2.

  In Patent Document 1, a plurality of fixed waveforms are prepared, and driving sound sources are generated by arranging these fixed waveforms at algebraically encoded sound source positions. With this configuration, high-quality output audio is obtained.

  In Non-Patent Document 2, a configuration in which a pitch filter is included in a generation unit of a driving sound source (ACELP sound source in Non-Patent Document 2) is being studied. The introduction of the fixed waveform and the pitch filter processing are simultaneously performed in the impulse response calculation part in Non-Patent Document 1, so that a quality improvement effect can be obtained without greatly increasing the search processing amount.

  Patent Document 2 discloses a configuration for searching for a pulse position while orthogonalizing a driving sound source to an adaptive sound source when the pitch gain is equal to or greater than a predetermined value.

  FIG. 17 is a block diagram showing a detailed configuration of the driving excitation encoding means 5 in the conventional CELP speech encoding apparatus in which the improved configurations of Patent Document 1 and Non-Patent Document 2 are introduced. In the figure, 16 is an auditory weighting filter coefficient calculating means, 17 and 19 are auditory weighting filters, 18 is a basic response generating means, 20 is a pre-table calculating means, 21 is a searching means, and 22 is a sound source position table.

Next, the operation of the drive excitation encoding means 5 will be described.
First, quantized linear prediction coefficients are input to the perceptual weighting filter coefficient calculation means 16 and the basic response generation means 18 from the linear prediction coefficient encoding means 3 in the speech encoding apparatus shown in FIG. The encoding target signal, which is the input speech 1 or the signal obtained by subtracting the synthesized sound from the adaptive excitation from the input speech 1, is input to the auditory weighting filter 17 from the means 4, and the adaptive excitation code is converted from the adaptive excitation encoding means 4. The repetition cycle of the adaptive sound source obtained in this way is input to the basic response generation means 18.

  The perceptual weighting filter coefficient calculating means 16 calculates perceptual weighting filter coefficients using the quantized linear prediction coefficients, and sets the perceptual weighting filter coefficients as the perceptual weighting filter 17 and perceptual weighting filter 19 filter coefficients. . The perceptual weighting filter 17 performs a filtering process on the input encoding target signal with the filter coefficient set by the perceptual weighting filter coefficient calculation means 16.

  The basic response generation means 18 performs a periodic process using a repetition period of the input adaptive sound source on a unit impulse or a fixed waveform, and uses the obtained signal as a sound source to generate the quantized linear prediction coefficient. A synthesized sound is generated by a synthesis filter constituted by using and is output as a basic response. The auditory weighting filter 19 performs a filtering process on the basic response using the filter coefficient set by the auditory weighting filter coefficient calculation unit 16.

  The pre-table calculation means 20 calculates a correlation value between the perceptually weighted encoding target signal and the perceptually weighted basic response as d (x), and calculates a percorrelation value of the perceptually weighted basic response. Let φ (x, y). Then, d ′ (x) and φ ′ (x, y) are obtained from the above equations (6) and (7), and these are stored as a pre-table.

  The sound source position table 22 stores sound source position candidates similar to those in FIG. The search means 21 sequentially reads out the sound source position candidates from the sound source position table 22 and pre-determines the evaluation value D for each combination of sound source positions based on the above formulas (1), (4), and (5). Calculation is performed using the pre-table calculated by the table calculation means 20. Then, the search means 21 searches for a combination of sound source positions that maximizes the evaluation value D, and uses the obtained sound source position codes (index in the sound source position table) and polarity representing the plurality of sound source positions as drive sound source codes. 14 and the time series vector corresponding to the driving excitation code are output to the gain encoding means 6 as a driving excitation.

  The introduction of orthogonalization disclosed in Patent Document 2 is performed by orthogonalizing the perceptually weighted encoding target signal input to the pre-table calculation means 20 with respect to the adaptive sound source and the above ( This is realized by subtracting the contribution related to the correlation between the adaptive sound source and each drive sound source from the value of E expressed by equation (5).

JP-A-10-232696 Japanese Patent Laid-Open No. 10-312198 Akitoshi Kataoka, Shinji Hayashi, Takehiro Moriya, Shoko Kurihara, Kazunori Mano “CS-ACELP Basic Algorithm” NTT R & D, Vol. 45, pp. 325-330, April 1996 Tsuchiya, Amada, and Mitseki, “Improvement of Adaptive Pulse Position ACELP Speech Coding” The Acoustical Society of Japan, 1999 Spring Research Conference Lecture I, pp. 213-214

  Since the conventional speech coding apparatus and speech decoding apparatus are configured as described above, the pitch periodization processing of the driving sound source can improve the coding characteristics without greatly increasing the search calculation processing amount. However, since the repetition period of the adaptive sound source is used as the repetition period used for periodicization, there is a problem that quality degradation occurs when the original pitch period is different from this repetition period.

  18 and 19 are diagrams for explaining the relationship between the encoding target signal and the sound source position of the periodic drive sound source in the conventional speech encoding device and speech decoding device. FIG. 18 shows a case where the repetition period of the adaptive sound source is about twice the original pitch period, and FIG. 19 shows a case where the repetition period of the adaptive sound source is about ½ times the original pitch period.

  Since the repetition period of the adaptive sound source is determined so as to minimize the coding distortion with respect to the signal to be coded, it is frequently a value different from the pitch period, which is the vibration period of the vocal cords. If they are different, they generally take a value that is an integer or an integral multiple of the original pitch period, and the most common values are 1/2 and 2 times.

  In FIG. 18, the vocal cord vibration periodically fluctuates every other pitch, so that the repetition period of the adaptive sound source is about twice the original pitch period. For this reason, when the driving sound source is encoded using this repetition period, the sound source positions are gathered in the first repetition period, and the result of repeating this in the repetition period in the frame is as shown in the figure. If a sound source repeated at a period different from the original pitch period is used, the timbre of the frame changes, resulting in an unstable impression of the synthesized sound. This problem cannot be ignored as the amount of sound source information of the driving sound source decreases as the bit rate is reduced, and becomes more pronounced when the amplitude of the adaptive sound source is smaller than the amplitude of the driving sound source.

  In FIG. 19, the low frequency component is dominant, and the waveforms of the first half and the latter half in the original pitch period are similar to each other, so that the repetition period of the adaptive sound source is about ½ times the original pitch period. I'm stuck. Also in this case, similarly to FIG. 18, since the sound source repeated at a cycle different from the original pitch cycle is used, the tone color of the frame changes and an unstable impression is generated in the synthesized sound.

  In addition, when the bit rate is reduced and the amount of information of the driving sound source is small, in the driving sound source determined to minimize the waveform distortion (coding distortion), the error in the low-amplitude band increases and Spectral distortion tends to increase, and this spectral distortion may be detected as sound quality degradation. Auditory weighting processing has been introduced to suppress the sound quality degradation due to this spectral distortion, but as the auditory weighting is increased, the waveform distortion increases and this causes a rough sound quality degradation. Adjustments are made so that the effects of sound quality degradation due to distortion and spectral distortion are comparable. However, the increase in the spectral distortion of the former is particularly large for female voices, and there is a problem that auditory weighting cannot be adjusted so as to be optimal for both male voices and female voices.

  In the conventional configuration, a constant amplitude is given within a frame to sound sources (including pulses) arranged at a plurality of sound source positions. When the number of candidates for each sound source position is compared, it is wasteful that the amplitude is constant despite the difference in the number. For example, in the sound source position table shown in FIG. 16, 3 bits are used for the sound source positions of sound source number 1 to sound source number 3, and 4 bits are used for the sound source position of sound source number 4. . When the maximum value of the correlation between the sound source and the encoding target signal at each position candidate is examined for each sound source number, it is easily predicted that the sound source number 4 with the largest number of candidates will obtain the largest probability value. The Considering an extreme case, consider a case where only 0 bits are given to a certain sound source number. When the sound source is arranged at 0 bits, that is, at a fixed position, even if the polarity is given separately, the correlation value is small, that is, it is not optimal to give a much larger amplitude than those of other sound source numbers. Therefore, there is a problem that the conventional configuration is not optimally designed with respect to amplitude.

  As for the amplitude for each sound source number, a configuration in which an independent value is separately given by vector quantization at the time of gain quantization is separately disclosed. However, this increases the amount of gain quantization information, makes the processing complicated, etc. There was a problem.

  Further, the introduction of orthogonalization of the driving sound source to the adaptive sound source has a configuration accompanied by an increase in search processing, and there has been a problem that when the number of combinations of algebraic sound sources increases, a large burden is imposed. In particular, when orthogonalization is performed in a configuration in which a fixed waveform or pitch period is introduced, there is a problem that the increase in the amount of calculation becomes even greater.

  The present invention has been made to solve the above-described problems, and an object thereof is to obtain a high-quality speech encoding apparatus and speech encoding method. Another object of the present invention is to obtain a high-quality speech encoding apparatus and speech encoding method while minimizing an increase in the amount of computation.

The speech coding apparatus according to the present invention includes a first impulse response when an impulse is set at a first pulse position among a plurality of pulse positions, and an impulse at a second pulse position among the plurality of pulse positions. Correlation value calculating means for calculating the first correlation value with the second impulse response at the time of setting for all combinations of the first pulse position and the second pulse position, and among the plurality of pulse positions A second correlation value between the impulse response when the impulse is set at one pulse position and the synthesized sound based on the adaptive sound source is calculated, and the first correlation value is corrected using the second correlation value. Correlation value correcting means; and a search means for calculating an evaluation value related to coding distortion using the corrected first correlation value and searching for a predetermined number of pulse positions of the driving sound source using the evaluation value; Correlation value correction means is based on adaptive sound source By using the power and the product of the second correlation value Naruoto it is obtained to perform the correction.

The speech encoding method according to the present invention includes a first impulse response when an impulse is set at a first pulse position among a plurality of pulse positions, and an impulse at a second pulse position among the plurality of pulse positions. A correlation value calculating step for calculating a first correlation value with the second impulse response at the time of setting for all combinations of the first pulse position and the second pulse position; A second correlation value between the impulse response when the impulse is set at one pulse position and the synthesized sound based on the adaptive sound source is calculated, and the first correlation value is corrected using the second correlation value. A correlation value correcting step; and a search step of calculating an evaluation value related to coding distortion using the corrected first correlation value and searching for a predetermined number of pulse positions of the driving sound source using the evaluation value, Correlation value correction step By using the power and the product of the second correlation value of the combined sound based on応音source is obtained to perform the correction.

According to the present invention, when the impulse is set at the first pulse position among the plurality of pulse positions, and when the impulse is set at the second pulse position among the plurality of pulse positions. Correlation value calculating means for calculating the first correlation value with the second impulse response for all combinations of the first pulse position and the second pulse position, and one pulse position among the plurality of pulse positions Correlation value correction means for calculating a second correlation value between the impulse response when the impulse is set up and the synthesized sound based on the adaptive sound source, and correcting the first correlation value using the second correlation value And a search means for calculating an evaluation value related to the coding distortion using the corrected first correlation value and searching for a predetermined number of pulse positions of the driving sound source using the evaluation value, and a correlation value correction means Is the power of synthesized sound based on adaptive sound source By performing correction using the product of the second correlation value, without increasing the amount of processing in the search unit, it can be orthogonalized with respect to the adaptive excitation encoding target signal, this by coding characteristics There is an effect that it can be improved and a high-quality speech encoding apparatus can be provided.

According to the present invention, when the impulse is set at the first pulse position among the plurality of pulse positions, and when the impulse is set at the second pulse position among the plurality of pulse positions. A correlation value calculating step for calculating a first correlation value with the second impulse response for all combinations of the first pulse position and the second pulse position, and one pulse position of the plurality of pulse positions A correlation value correction step of calculating a second correlation value between the impulse response when the impulse is raised and the synthesized sound based on the adaptive sound source, and correcting the first correlation value using the second correlation value And a search step for calculating an evaluation value related to coding distortion using the corrected first correlation value and searching for a predetermined number of pulse positions of the driving sound source using the evaluation value, and a correlation value correction step Is based on adaptive sound source By performing correction using the power and the product of the second correlation value of the synthesized sound, without increasing the amount of processing in the search step, can be orthogonalized with respect to the adaptive excitation encoding target signal, which Thus, the encoding characteristics can be improved, and a high-quality speech encoding method can be provided.

Hereinafter, an embodiment of the present invention will be described.
Embodiment 1 FIG.
FIG. 1 is a block diagram showing a configuration of driving excitation encoding means 5 in the speech encoding apparatus according to Embodiment 1 of the present invention. The overall configuration of the speech encoding apparatus is the same as that shown in FIG. In the figure, reference numeral 23 is a period preliminary selection means, 27 is a driving excitation encoding means, 28 is a period encoding means, and the period preliminary selection means 23 comprises a constant table 24, a comparison means 25, and a preliminary selection means 26. Yes.

  The driving excitation encoding unit 27 is a unit that operates in the same manner as the conventional driving excitation encoding unit 5. However, the period preliminary selection unit 23 and the period encoding unit 28 are provided before and after the driving excitation encoding unit 27. What is newly added as a part of the drive excitation encoding means 5 in FIG. 14 is the speech encoding apparatus according to the first embodiment.

  FIG. 2 is a block diagram showing a configuration of driving excitation decoding means 12 in the speech decoding apparatus according to Embodiment 1 of the present invention. The overall configuration of the speech decoding apparatus is the same as that shown in FIG. In FIG. 2, 29 is a cyclic decoding means, and 30 is a driving excitation decoding means.

  The driving excitation decoding unit 30 is a unit that operates in the same manner as the conventional driving excitation decoding unit 12, but the periodic preliminary selection unit 23 and the periodic decoding unit 29 are provided before the driving excitation decoding unit 30. What is newly inserted is a part of the driving excitation decoding means 12 in FIG. 15 is the speech decoding apparatus according to the first embodiment.

Next, the operation will be described.
First, the operation of the speech encoding apparatus will be described with reference to FIG. The adaptive excitation repetition period obtained by converting the adaptive excitation code is input from the adaptive excitation encoding means 4 shown in FIG. The encoding target signal from the adaptive excitation encoding unit 4 and the quantized linear prediction coefficient from the linear prediction coefficient encoding unit 3 are input to the driving excitation encoding unit 27.

  The constant table 24 in the period preliminary selection means 23 stores three constants 1/2, 1, and 2. Each constant is multiplied by the repetition period of the input adaptive sound source, and the obtained three constants. The repetition period is output to the preliminary selection means 26 as a repetition period candidate of the driving sound source. The comparison unit 25 compares the repetition period of the input adaptive sound source with a predetermined threshold value that is given in advance, and outputs the comparison result to the preliminary selection unit 26. As the predetermined threshold, about 40 corresponding to an average pitch period is used.

  When the comparison result from the comparison unit 25 exceeds the predetermined threshold, the preliminary selection unit 26 repeats the repetition cycle of two driving sound sources by multiplying the repetition cycle of the input adaptive sound source by 1/2 or 1. When a candidate is preliminarily selected and the comparison result is a result equal to or less than a predetermined threshold value, two repetition periods candidates of driving sound sources obtained by multiplying the input adaptive sound source repetition period by 1, 2 are preliminarily selected and obtained. The repetition period candidates of the two driving excitations are sequentially output to the driving excitation encoding means 27.

  Similarly to the conventional driving excitation encoding means 5 shown in FIG. 17, the driving excitation encoding means 27 is a repetition period candidate of two input driving excitations (the difference from FIG. 17 is that this repetition period is an adaptive excitation excitation). Is a constant multiple of), and the quantized linear prediction coefficient and the encoding target signal are used to perform encoding processing of the algebraic excitation, and for each repetition period candidate of the two driving excitations, The evaluation value D in the above equation (1) relating to the sound source position, polarity and coding distortion at that time that minimizes the distortion is output.

  The period encoding unit 28 compares the evaluation values D for the repetition period candidates of each driving excitation output from the driving excitation encoding unit 27, and the difference between one evaluation value and the remaining evaluation values is equal to or greater than a predetermined threshold value. (That is, only one of them has a small encoding distortion), the repetition period candidate of the driving excitation that gave the evaluation value is selected, and when the difference between the evaluation values is less than a predetermined threshold, Select the repetition period candidate of the driving sound source closest to the pitch period (original pitch period estimation result) that has been analyzed separately, and select information obtained by encoding this selection result with 1 bit, and the sound source position at that time The sound source position code and polarity to be expressed are output to the multiplexing means 7 shown in FIG. 14 as a drive sound source code, and the time-series vector corresponding to this drive sound source code is used as the drive sound source and the gain encoding means 6 shown in FIG. Out To.

  Next, the operation of the speech decoding apparatus will be described with reference to FIG. In the speech decoding apparatus shown in FIG. 15, the separation unit 9 separates the speech code 8 output from the speech encoding apparatus and converts the code of the linear prediction coefficient to the linear prediction coefficient decoding unit 10 as in the conventional case. Output, the adaptive excitation code is output to the adaptive excitation decoding means 11, the driving excitation code is output to the driving excitation decoding means 12, and the gain code is output to the gain decoding means 13. In this embodiment, The adaptive excitation repetition period obtained by converting the adaptive excitation code is input to the driving excitation decoding means 12 from the adaptive excitation decoding means 11 shown in FIG. That is, in FIG. 2, the adaptive excitation repetition period is input from the adaptive excitation decoding means 11 to the period preliminary selection means 23. The selection information in the driving excitation code separated by the separating means 9 is input to the periodic decoding means 29, and the excitation position code and polarity in the driving excitation code are input to the driving excitation decoding means 30.

  The period preliminary selection means 23 has the same configuration as the period preliminary selection means 23 shown in FIG. 1 in the speech coding apparatus, and the preliminary selection means 26 includes a plurality of drive excitation sources obtained by multiplying the repetition period of the input adaptive excitation by a constant. Based on the comparison result of the comparison means 25, the repetition period candidates of the two preselected driving excitations are selected from the repetition period candidates and output to the period decoding means 29.

  The period decoding means 29 selects one of the two repetition periods of the drive sound source selected from the preliminary selection means 26 output from the preliminary selection means 26 according to the input selection information, and uses this as the drive sound source repetition period. The data is output to the decryption means 30. Similarly to the conventional drive excitation decoding unit 12, the drive excitation decoding unit 30 arranges a fixed waveform at each position corresponding to the excitation position code, performs a pitch period based on the repetition period, and generates a drive excitation code. The corresponding time series vector is output as a driving sound source.

  3 and 4 are diagrams for explaining the relationship between the encoding target signal and the sound source position of the periodic drive sound source in the speech coding apparatus and speech decoding apparatus according to Embodiment 1. FIG. The encoding target signal is the same as that shown in FIGS. 18 and 19, and FIG. 3 shows a case where the repetition period of the adaptive excitation is about twice the original pitch period, and FIG. This is the case.

  In the case of FIG. 3, if the original pitch period is 20 or more, the adaptive sound source repetition period is 40 or more. Therefore, in the preliminary selection means 26, in most cases, 1/2 times the repetition period of the adaptive sound source and 1 A double value is preselected. If the difference in the evaluation value D at the time of encoding when using these two repetition periods is small, it is close to the estimated value of the original pitch period (the accuracy rate is higher than the repetition period of the adaptive excitation) that is separately obtained. / 2 times is selected, and ideally periodic sound source positions are obtained as shown in the figure.

  In the case of FIG. 4, if the original pitch period is less than 80, the repetition period of the adaptive sound source is less than 40. Therefore, the preliminary selection means 26 preliminarily selects values of 1 and 2 times that of the adaptive sound source with high probability. Is done. If the difference between the evaluation values D at the time of encoding when using these two repetition periods is small, a value close to the original pitch period that is separately obtained is selected, and ideally periodic as shown in the figure. Sound source position is obtained.

  In the above embodiment, the algebraic sound source expressed only by the positions and polarities of several pulses is used for encoding and decoding of the drive sound source. However, the present invention is limited to the algebraic sound source configuration. However, the present invention can also be applied to CELP speech encoding apparatuses and speech decoding apparatuses that use other learning excitation codebooks, random excitation codebooks, and the like.

  In the above embodiment, the pitch period is separately obtained and used for selection by the period encoding means 28. However, the encoding distortion is minimized without using this, that is, the evaluation value D is maximized. A configuration in which the repetition period is selected is also possible. Further, instead of the pitch period, a value obtained by averaging the repetition periods of the adaptive sound sources in the past several frames may be used as the reference value.

  Further, in the above-described embodiment, the linear prediction coefficient is used as the spectral parameter. However, a configuration using other spectral parameters such as LSP (Line Spectrum Pair) generally used may be used.

  Furthermore, in the above embodiment, all constants in the constant table 24 are multiplied by the repetition period of the adaptive sound source. However, the preliminary selection means 26 selects two constants from the constant table 24, and then the adaptive sound source. The same applies if the repetition period is multiplied.

  Further, the same result can be obtained by deleting 1 from the constant table and inputting the repetition period of the adaptive sound source directly to the preliminary selection means 26 instead.

  Furthermore, although the characteristic improvement effect is reduced, a configuration in which the comparison unit 25 and the preliminary selection unit 26 are eliminated by setting the values in the constant table to only 1/2 and 1 is possible.

  As described above, according to the first embodiment, a repetition cycle candidate of a plurality of driving sound sources is obtained by multiplying a repetition cycle of the adaptive sound source by a plurality of constants, and a predetermined repetition cycle candidate of the plurality of driving sound sources is determined. Based on the result of searching for a driving excitation code that minimizes the encoding distortion for each repetition period candidate of the preselected driving excitation and comparing the encoding distortion for each repetition period of the driving excitation. Therefore, even if the original pitch period and the adaptive sound source repetition period are different, the drive sound source is cycled using a repetition period close to the original pitch period with a high probability. Is selected, it is possible to suppress the generation of an unstable impression of the synthesized sound and to obtain an effect that a high-quality speech encoding apparatus can be provided.

  In addition, since the number of preliminary selections in the period preliminary selection is set to 2 and the selection information of the repetition period of the driving sound source is encoded with 1 bit, a high-quality speech encoding apparatus is provided by adding a minimum amount of information. The effect that it can be obtained.

  Furthermore, in the cycle preliminary selection, the repetition cycle of the adaptive sound source is compared with a predetermined threshold value, and a predetermined number of drive sound source repetition cycle candidates are selected based on the comparison result. Low-probability driving excitation repetition period candidates can be eliminated, driving excitation encoding processing and allocation of selection information for driving excitation repetition period candidates that do not need to be evaluated become unnecessary, and a minimum amount of computation and information are added Thus, it is possible to provide a high-quality speech encoding apparatus.

  Furthermore, since the constants to multiply the repetition period of the adaptive sound source in the period preselection are included at least 1/2, 1, the repetition period candidate of the driving sound source including the original pitch period is selected with a high probability with few options. Thus, an effect that a high-quality speech encoding apparatus can be provided with the addition of the minimum amount of calculation and the amount of information can be obtained.

  Furthermore, according to the first embodiment, a repetition period candidate for a plurality of driving sound sources is obtained by multiplying the repetition period of the adaptive sound source by a plurality of constants, and a predetermined number is reserved among the repetition period candidates for the plurality of driving sound sources. Based on the selection information of the repetition period of the driving sound source in the voice code, one of the repetition period candidates of the pre-selected driving sound source is selected as the repetition period of the driving sound source, and the repetition period of this driving sound source Therefore, even if the original pitch period and the repetition period of the adaptive sound source are different, the drive sound source is periodicized using a repetition period close to the original pitch period with high probability. Thus, it is possible to suppress the generation of an unstable impression of the synthesized sound, and to obtain an effect that a high-quality speech decoding apparatus can be provided.

  Furthermore, since the number of preliminary selections in the periodical preliminary selection is 2, and the selection information of the repetition period of the driving sound source encoded with 1 bit is decoded, high-quality speech decoding can be performed with the addition of a minimum amount of information. The effect that the control apparatus can be provided is obtained.

  Furthermore, in the cycle preliminary selection, the repetition cycle of the adaptive sound source is compared with a predetermined threshold value, and a predetermined number of drive sound source repetition cycle candidates are selected based on the comparison result. Low-probability driving sound source repetition period candidates can be eliminated, selection information is not distributed to unnecessary driving sound source repetition period candidates, and a high-quality speech decoding apparatus can be provided by adding a minimum amount of information. The effect is obtained.

  Furthermore, since the constants to multiply the repetition period of the adaptive sound source in the period preselection are included at least 1/2, 1, the repetition period candidate of the driving sound source including the original pitch period is selected with a high probability with few options. Thus, an effect that a high-quality speech decoding apparatus can be provided with the addition of a minimum amount of information can be obtained.

Embodiment 2. FIG.
FIG. 5 is a block diagram showing a configuration of driving excitation encoding means 5 in the speech encoding apparatus according to Embodiment 2 of the present invention. The overall configuration of the speech encoding apparatus is the same as that of the first embodiment, that is, FIG. In FIG. 5, 31 is a periodic preliminary selection means, 33 is an adaptive excitation codebook stored in the adaptive excitation encoding means 4, and the periodic preliminary selection means 31 includes a constant table 32, an adaptive excitation generation means 34, a distance. The calculation means 35 and the preliminary selection means 36 are configured.

  The driving excitation encoding unit 27 is a unit that operates in the same manner as the conventional driving excitation encoding unit 5, but the period preliminary selection unit 31 and the period encoding unit 28 are provided before and after the driving excitation encoding unit 27. What is newly inserted is a part of the drive excitation encoding means 5 in FIG. 14 is the speech encoding apparatus according to the second embodiment.

  FIG. 6 is a block diagram showing the configuration of the driving excitation decoding means 12 in the speech decoding apparatus according to Embodiment 2 of the present invention. The overall configuration of the speech decoding apparatus is the same as that of Embodiment 1, that is, FIG. In FIG. 6, reference numeral 33 denotes an adaptive excitation codebook stored in the adaptive excitation decoding means 11.

  The driving excitation decoding unit 30 is a unit that operates in the same manner as the conventional driving excitation decoding unit 12. However, the periodic preliminary selection unit 31 and the periodic decoding unit 29 are provided before the driving excitation decoding unit 30. What is newly inserted is a part of the drive excitation decoding means 12 in FIG. 15 is the speech decoding apparatus according to the second embodiment.

Next, the operation will be described.
First, the operation of the speech encoding apparatus will be described with reference to FIG. Similar to the first embodiment, the repetition period of the adaptive excitation output from the adaptive excitation encoding unit 4 is input to the period preliminary selection unit 31, and the encoding target signal and the linear prediction coefficient code from the adaptive excitation encoding unit 4 are input. The quantized linear prediction coefficient from the encoding unit 3 is input to the driving excitation encoding unit 27.

  Four constants 1/3, 1/2, 1, and 2 are stored in the constant table 32 in the period preliminary selection means 31, and each constant is multiplied by the repetition period of the input adaptive sound source. The four drive sound source repetition period candidates thus output are output to the adaptive sound source generation means 34 and the preliminary selection means 36.

  The adaptive excitation generator 34 generates an adaptive excitation using the past excitations stored in the adaptive excitation codebook 33, with each of the four drive excitation repetition period candidates as a repetition period, The generated four adaptive sound sources are output to the distance calculation means 35. It should be noted that, for a value that is one time the repetition period of the adaptive excitation, the adaptive excitation encoding means 4 has already generated the same adaptive excitation, and therefore generation by the adaptive excitation generation means 34 can be omitted. .

  In addition, when some of the repetition period candidates of the four driving excitations are too large or too small to have an inappropriate value as the pitch period, the adaptive excitation codebook 33 may not be able to cope with it. Since this may occur, the adaptive sound source generation means 34 outputs a 0 signal as an adaptive sound source for the drive sound source repetition period candidate so that it is not selected during the subsequent preliminary selection.

  The distance calculation means 35 is an adaptive excitation (that is, the adaptive excitation output from the adaptive excitation encoding means 4) when the repetition period is set to a value that is one time the repetition period of the adaptive excitation, the other 1/3 times, 1 / The distance between the adaptive sound source when the double and double values are used as the repetition period is calculated, and each obtained distance is output to the preliminary selection means 36.

  The preliminary selection means 36 first compares the distances at 1/3 times and 1/2 times and selects the smaller one. Then, the selected distance is compared with a value obtained by multiplying the average amplitude of the adaptive sound source by a predetermined constant, and when the former is small, the repetition period (1/3 times or 1 times the repetition period of the adaptive sound source) is given. / 2) and a value one times the repetition period of the adaptive sound source are output as the repetition period candidates of the preselected drive sound source. When the former is greater than or equal to the latter, the distance is then compared with the distance at twice the repetition period of the adaptive sound source, and the repetition period giving the smaller distance and the value of one time of the repetition period of the adaptive sound source, Output as a repetition cycle candidate of the pre-selected drive sound source. The predetermined constant may be a positive value less than 1 and a small value of about 0.1.

  Similarly to the conventional driving excitation encoding means 5 shown in FIG. 17, the driving excitation encoding means 27 is a repetition period candidate for each inputted preselected driving excitation (the difference from FIG. 17 is this preliminary selection). The repetition period candidate of the generated driving sound source is a constant multiple of the adaptive sound source), the quantized linear prediction coefficient, and the encoding target signal are used to perform algebraic sound source encoding processing, A drive excitation code that minimizes the coding distortion is searched for each repetition candidate, and the obtained plurality of sound source positions and polarities and the evaluation value D of the above equation (1) relating to the coding distortion at that time are output.

  The period encoding means 28 compares the evaluation values for the repetition period candidates of the driving excitation output from the driving excitation encoding means 27, and the difference between one evaluation value and the remaining evaluation values is equal to or greater than the threshold value ( In other words, if only one of them has a small coding distortion), the repetition period candidate of the driving sound source to which the evaluation value is given is selected, and if the difference between the evaluation values is less than the threshold value, it is analyzed separately. Select the repetition period candidate of the driving sound source closest to the pitch period (the original pitch period estimation result), select information obtained by encoding this selection result with 1 bit, and the sound source position code and polarity indicating the sound source position at that time Are output as drive excitation codes.

  Next, the operation of the speech decoding apparatus will be described with reference to FIG. As in the first embodiment, the repetition period of the adaptive excitation output from the adaptive excitation decoding means 11 is input to the periodic preliminary selection means 31, and the selection information in the driving excitation code separated by the separation means 9 is the period decoding means. 29, and the excitation position code and polarity in the driving excitation code are input to the driving excitation decoding means 30.

  The period preliminary selection means 31 has the same configuration as the period preliminary selection means 31 shown in FIG. 5 in the speech coding apparatus, and two preliminary period candidates for driving excitation repetition periods obtained by multiplying the repetition period of the input adaptive excitation by a constant number. A repetition cycle candidate of the selected driving sound source is selected and output to the cyclic decoding means 29. Periodic decoding means 29 selects one of the two driving excitation repetition period candidates according to the inputted driving excitation selection information, and outputs this to driving excitation decoding means 30 as the repetition period of the driving excitation. Similarly to the conventional drive excitation decoding unit 12, the drive excitation decoding unit 30 arranges a fixed waveform at each position corresponding to the excitation position code, performs pitch cycle based on the repetition period, A time series vector is output as a driving sound source.

  7, 8, and 9 are diagrams for explaining the adaptive excitation generated by the adaptive excitation generation means 34 in the speech encoding apparatus and speech decoding apparatus according to Embodiment 2, and FIG. FIG. 8 shows the case where the repetition period of the adaptive sound source is twice the original pitch period, and FIG. 9 shows the case where the repetition period of the adaptive sound source is the original pitch period. This is a case of 3 times.

  Referring to FIG. 7, when the repetition cycle of the adaptive sound source matches the original pitch cycle, the adaptive sound source generated by using 1/3 times and 1/2 times the repetition cycle of the adaptive sound source as the repetition cycle It can be seen that the distance from the adaptive sound source (the uppermost one in the figure) is large and 2 and 1 times are likely to be pre-selected.

  Referring to FIG. 8, when the repetition period of the adaptive sound source is twice the original pitch period, the adaptive sound source generated with a repetition period of ½ times the repetition period of the adaptive sound source and the original adaptive sound source (see FIG. It can be seen that the distance from the uppermost one is small, and 1/2 and 1 times are likely to be preliminarily selected.

  Referring to FIG. 9, when the repetition cycle of the adaptive sound source is three times the original pitch cycle, the adaptive sound source generated by using 1/3 times the repetition cycle of the adaptive sound source as the repetition cycle and the original adaptive sound source (see FIG. 9). It can be seen that the distance to the topmost one is small, and 1/3 and 1 times are likely to be preselected.

  In the above embodiment, an algebraic excitation is used for encoding and decoding of the driving excitation, but the present invention is not limited to the algebraic excitation configuration, and other learning excitation codebooks and random The present invention can also be applied to a CELP speech encoding apparatus and speech decoding apparatus that use a sound source codebook or the like.

  In the above-described embodiment, the pitch period is separately obtained and used for selection by the period encoding means 28. However, without using this, the encoding distortion is minimized, that is, the evaluation value D is maximized. A configuration for selecting a repetition period candidate of a sound source is also possible. Instead of the pitch period, a value obtained by averaging the repetition periods of the adaptive sound sources of the past several frames may be used as the reference value.

  Furthermore, in the above-described embodiment, the linear prediction coefficient is used as the spectrum parameter. However, a configuration using other spectrum parameters such as LSP that is generally used may be used.

  Furthermore, the same result can be obtained by deleting 1 from the constant table and inputting the repetition period of the adaptive sound source directly to the preliminary selection means 36 instead.

  Furthermore, although the characteristic improvement effect is reduced, a configuration in which the values in the constant table are only 1/2, 1, 2 is possible.

  As described above, according to the second embodiment, a repetition cycle candidate for a plurality of driving sound sources is obtained by multiplying a repetition cycle of the adaptive sound source by a plurality of constants, and the repetition cycle candidates for the plurality of driving sound sources are applied as they are. Each adaptive sound source is generated when the repetition period of the sound source is set, and a repetition period candidate of a predetermined number of driving sound sources is selected based on the distance value between the generated adaptive sound sources. Even when the repetition period of the adaptive sound source is different, the drive sound source periodicity is selected with a repetition period close to the original pitch period with high probability, and the generation of unstable impressions of the synthesized sound can be suppressed, resulting in high-quality sound. The effect that an encoding device can be provided is obtained.

  Furthermore, since the number of preliminary selections in the period preliminary selection is set to 2, and the selection information of the repetition period of the driving sound source is encoded with 1 bit, a high-quality speech encoding apparatus can be provided by adding a minimum amount of information. The effect that it can be obtained.

  Further, each of the adaptive sound sources is generated when the repetition period candidates of the plurality of driving sound sources are directly used as the repetition period of the adaptive sound source, and based on the generated distance value between the adaptive sound sources, the repetition period of the predetermined number of driving sound sources Since the candidate is selected, it is possible to eliminate the repetition period candidate of the driving excitation with a low probability that it is the original pitch period, and the driving excitation encoding process and the distribution of the selection information for the repetition period candidate of the driving excitation that does not need to be evaluated Is not required, and an effect that a high-quality speech coding apparatus can be provided by adding a minimum amount of calculation and an amount of information can be obtained.

  In addition, since the constant to multiply the repetition period of the adaptive sound source in the period preselection is included at least 1/2, 1, the repetition period candidate of the driving sound source including the original pitch period is generated with high probability with few options. Thus, an effect that a high-quality speech encoding apparatus can be provided with the addition of the minimum amount of calculation and the amount of information can be obtained.

  Further, according to the second embodiment, a repetition cycle candidate of a plurality of driving sound sources is obtained by multiplying a repetition cycle of the adaptive sound source by a plurality of constants, and a predetermined number of spares are selected from the repetition cycle candidates of the plurality of driving sound sources. A repetition period candidate of the selected driving sound source is selected, and one of the repetition period candidates of the pre-selected driving sound source is selected based on the selection information of the repetition period of the driving sound source in the speech code. Since the driving sound source is decoded using this repetition period, even when the original pitch period and the repetition period of the adaptive sound source are different, the repetition period close to the original pitch period is used with a high probability. The driving sound source is periodicized, the generation of an unstable impression of the synthesized sound can be suppressed, and an effect that a high-quality speech decoding apparatus can be provided can be obtained.

  Furthermore, since the number of preliminary selections in the periodical preliminary selection is 2, and the selection information of the repetition period of the driving sound source encoded with 1 bit is decoded, high-quality speech decoding can be performed with the addition of a minimum amount of information. The effect that the control apparatus can be provided is obtained.

  Further, in the period preselection, each of the adaptive sound sources when the repetition period candidates of the plurality of driving sound sources are directly used as the repetition period of the adaptive sound source is generated, and based on the distance value between the generated adaptive sound sources, a predetermined number of Since the repetition period candidate of the driving sound source is selected, it is possible to eliminate the repetition period candidate of the driving sound source that has a low probability of being the original pitch period, and it is not necessary to distribute selection information to the repetition period candidate of the unnecessary repeated driving sound source Thus, an effect that a high-quality speech decoding apparatus can be provided with the addition of a minimum amount of information can be obtained.

  Furthermore, since the constants to multiply the repetition period of the adaptive sound source in the period preselection are included at least 1/2, 1, the repetition period candidate of the driving sound source including the original pitch period is selected with a high probability with few options. Thus, an effect that a high-quality speech decoding apparatus can be provided with the addition of a minimum amount of information can be obtained.

Embodiment 3 FIG.
FIG. 10 is a block diagram showing the configuration of the drive excitation encoding means 5 and the newly added auditory weighting control means 37 in the speech encoding apparatus according to Embodiment 3 of the present invention. The overall configuration of the speech encoding apparatus is the same as that shown in FIG. 14 except that an auditory weighting control unit 37 is added to the driving excitation encoding unit 5. The auditory weighting control unit 37 includes a comparison unit 38 and an intensity control unit 39. The configuration in the driving excitation encoding means 5 is the same as that of the conventional one explained in FIG. 17, and only the point that the auditory weighting filter coefficient calculating means 16 is controlled by the auditory weighting control means 37 is changed. Yes.

Next, the operation will be described.
First, quantized linear prediction is performed from the linear prediction coefficient coding unit 3 shown in FIG. 14 in the speech coding apparatus to the perceptual weighting filter coefficient calculation unit 16 and the basic response generation unit 18 in the drive excitation coding unit 5. A coefficient is entered. The adaptive excitation repetition period obtained by converting the adaptive excitation code from the adaptive excitation encoding means 4 to the basic response generating means 18 in the driving excitation encoding means 5 and the comparison means 38 in the auditory weighting control means 37. Is entered. Further, the input signal 1 or a signal obtained by subtracting the synthesized sound from the adaptive sound source from the input sound 1 is input from the adaptive excitation encoding unit 4 to the auditory weighting filter 17 in the driving excitation encoding unit 5 as an encoding target signal. The

  The comparison means 38 in the auditory weighting control means 37 compares the inputted repetition period with a predetermined threshold value and outputs the comparison result to the intensity control means 39. The predetermined threshold is a value of about 40 that substantially separates the distribution of the pitch period of the male voice and female voice.

  The intensity control means 39 determines an intensity coefficient for controlling the enhancement intensity in the perceptual weighting filter based on the comparison result, and the determined intensity coefficient is supplied to the perceptual weighting filter coefficient calculation means 16 in the drive excitation encoding means 5. Output. In the comparison result of the comparison means 38, when the repetition period of the adaptive sound source is equal to or greater than a predetermined threshold, since the possibility of male voice is high, the intensity coefficient is determined so that the intensity of auditory weighting is weakened. If the repetition period of the adaptive sound source is less than the predetermined threshold in the opposite comparison result, the possibility of being a female voice is high, and the intensity coefficient is determined so that the intensity of auditory weighting is increased. The intensity coefficient is a multiplication value of the linear prediction coefficient used for calculation of the auditory weighting filter coefficient.

  The perceptual weighting filter coefficient calculating means 16 calculates perceptual weighting filter coefficients using the quantized linear prediction coefficient and the intensity coefficient, and the perceptual weighting filter coefficients are calculated by the perceptual weighting filter 17 and perceptual weighting filter 19. Set as a filter coefficient.

  The subsequent configurations and operations of the perceptual weighting filter 17, basic response generation means 18, perceptual weighting filter 19, pre-table calculation means 20, search means 21, and sound source position table 22 are the same as those in the prior art and will not be described.

  In the above embodiment, the intensity coefficient is determined based on whether the auditory weighting control means 37 is greater than or less than a predetermined threshold value. However, the intensity coefficient is more finely controlled using two or more predetermined threshold values, It is also possible to control continuously based on the magnitude of the difference.

  In the above embodiment, an algebraic excitation is used for encoding the driving excitation. However, the present invention is not limited to the algebraic excitation configuration, and other learning excitation codebooks and random excitation codebooks. The present invention can also be applied to a CELP speech coding apparatus using the above.

  Furthermore, in the above-described embodiment, the linear prediction coefficient is used as the spectrum parameter. However, a configuration using other spectrum parameters such as LSP that is generally used may be used.

  As described above, according to the third embodiment, the intensity coefficient of auditory weighting is controlled based on the value of the repetition period of the adaptive sound source, and the filter coefficient for auditory weighting is calculated using this intensity coefficient. Since this filter coefficient is used to perform auditory weighting on the signal to be encoded that encodes the driving sound source, it is possible to perform auditory weighting that is optimally adjusted for both male and female voices. The effect that the control apparatus can be provided is obtained.

Embodiment 4 FIG.
FIG. 11 is a block diagram showing the configuration of driving excitation encoding means 5 and newly added auditory weighting control means 40 in the speech encoding apparatus according to Embodiment 4 of the present invention. The overall configuration of the speech encoding apparatus is the same as that shown in FIG. 14 except that an auditory weighting control unit 40 is added to the driving excitation encoding unit 5. The auditory weighting control means 40 includes a comparison means 38, an intensity control means 39, and an average value update means 41. The configuration in the driving excitation encoding means 5 is the same as that of the conventional one explained in FIG. 17, and only the point that the auditory weighting filter coefficient calculating means 16 is controlled by the auditory weighting control means 40 is changed. Yes.

Next, the operation will be described.
Since the fourth embodiment has a configuration in which the average value updating means 41 is added to the auditory weighting control means 37 of the first embodiment, the operation of this new part will be mainly described. The adaptive excitation repetition period obtained by converting the adaptive excitation code from the adaptive excitation coding means 4 to the basic response generation means 18 in the driving excitation coding means 5 and the average value updating means 41 in the auditory weighting control means 40. Is entered.

  The average value updating unit 41 in the auditory weighting control unit 40 updates the average value of the repetition period of the adaptive sound source stored therein by using the input repetition period of the adaptive sound source, and compares the updated average value. Output to the means 38. The simplest method for updating the average value includes a method in which the repetition period of the frame is multiplied by a constant α smaller than 1 and a method in which the average value so far is multiplied by 1-α. The purpose of obtaining the average value is to stably determine whether the voice is male or female. Therefore, it is desirable to update after limiting the update to a frame having a large adaptive sound source gain.

  Then, the comparison unit 38 compares the updated average value with a predetermined threshold value and outputs the comparison result to the intensity control unit 39. The intensity control means 39 determines an intensity coefficient for controlling the enhancement intensity in the perceptual weighting filter based on the comparison result, and outputs the determined intensity coefficient to the perceptual weighting filter coefficient calculation means 16 in the drive excitation encoding means 5. To do. In the comparison result of the comparison means 38, when the average value is equal to or greater than a predetermined threshold value, there is a high possibility that the voice is male voice. In the opposite comparison result, when the average value is less than the predetermined threshold value, there is a high possibility that the voice is a female voice. Therefore, the intensity coefficient is determined so that the intensity of auditory weighting is increased.

  The configuration and operation of the subsequent auditory weighting filter coefficient calculating means 16, auditory weighting filter 17, basic response generating means 18, auditory weighting filter 19, pre-table calculating means 20, search means 21, and sound source position table 22 are the same as in the prior art. Since there is, description is abbreviate | omitted.

  In the above embodiment, the intensity coefficient is determined based on whether the auditory weighting control means 40 is greater than or less than a predetermined threshold. However, the intensity coefficient is more finely controlled using two or more predetermined thresholds, It is also possible to control continuously based on the magnitude of the difference from the threshold.

  In the above embodiment, an algebraic excitation is used for encoding the driving excitation. However, the present invention is not limited to the algebraic excitation configuration, and other learning excitation codebooks and random excitation codebooks. The present invention can also be applied to a CELP speech coding apparatus using the above.

  Furthermore, in the above-described embodiment, the linear prediction coefficient is used as the spectrum parameter. However, a configuration using other spectrum parameters such as LSP that is generally used may be used.

  As described above, according to the fourth embodiment, the intensity coefficient of auditory weighting is controlled based on the past average value of the repetition period of the adaptive sound source, and the filter coefficient for auditory weighting is used using this intensity coefficient. This filter coefficient is used to perform auditory weighting on the signal to be encoded that encodes the driving sound source, so that auditory weighting that is optimally adjusted for both male and female voices is possible, resulting in high quality. The effect of providing a speech encoding apparatus can be obtained.

  In particular, by using the past average value of the repetition cycle of the adaptive sound source, it is possible to suppress the occurrence of an unstable impression due to frequent changes in the intensity of auditory weighting.

Embodiment 5. FIG.
FIG. 12 is a diagram showing the excitation position table 22 used in the driving excitation encoding means 5 in the speech encoding apparatus and the driving excitation decoding means 12 in the speech decoding apparatus according to Embodiment 5 of the present invention. In the conventional sound source position table shown in FIG. 16, a fixed amplitude is added for each sound source number.

  The amplitude value of this fixed amplitude is given according to the number of sound source position candidates for each sound source number, if within the same table. In the case of FIG. 12, the number of sound source position candidates from sound source number 1 to sound source number 3 is 8, and the same amplitude value 1.0 is given. Since the sound source number 4 has as many as 16 sound source position candidates, an amplitude value 1.2 larger than the others is given. Thus, the larger the number of sound source position candidates, the larger the amplitude value is given.

ｄ”（ｍ_k ）＝ａ_k ｄ’（ｍ_k ） （１０）
φ”（ｍ_k ，ｍ_i ）＝ａ_k ａ_i φ’（ｍ_k ，ｍ_i ） （１１）
とする。ここで、ａ_k はｋ番目のパルスの振幅（図１２の振幅）である。パルス位置の全組合せに対する評価値Ｄの計算を始める前に、ｄ”とφ”の計算を行っておくことにより、後は（８）式と（９）式の単純加算という少ない演算量で評価値Ｄが算出できる。 The sound source position search using the sound source position table to which the amplitude is given can also be performed based on the above equation (1). However,

d ″ (m _k ) = a _k d ′ (m _k ) (10)
φ ″ (m _k , m _i ) = a _k a _i φ ′ (m _k , m _i ) (11)
And Here, a _k is the amplitude of the kth pulse (the amplitude in FIG. 12). By calculating d ″ and φ ″ before starting the calculation of the evaluation value D for all combinations of pulse positions, the subsequent evaluation is performed with a small amount of calculation, which is a simple addition of the equations (8) and (9). A value D can be calculated.

  The driving sound source is decoded based on the sound source position code by selecting one sound source position for each sound source number in the sound source position table of FIG. 12 and giving the sound source position for each sound source number. This is done by placing a sound source multiplied by a fixed amplitude. When the sound source is not a pulse or is periodic, the components of the sound source to be arranged overlap, so all the overlapping portions may be added. That is, in the conventional algebraic excitation decoding process, a process of multiplying the fixed amplitude given for each excitation number is added.

  In the prior art, a fixed waveform is prepared for each sound source number, but in that case, a basic response has to be calculated for each sound source number. In this embodiment, only the correction of the pre-table is added as described above. In the conventional technique, the amplitude value is not given in accordance with the difference in the amount of position information (number of candidates) depending on the sound source number.

  As described above, according to the fifth embodiment, a fixed amplitude is given in advance based on the number of selectable candidates for each sound source position, and the drive sound source encoding means 5 is a sound source arranged at the sound source position. When the drive sound source is generated by adding all the sound sources while multiplying by the fixed amplitude, the code and polarity representing the sound source position that gives the drive sound source with the least coding distortion with the input sound are searched and output. As a result, it is possible to obtain a high-quality speech coding apparatus with a simple configuration, with little increase in processing amount, and reduced waste regarding amplitude for each sound source.

  Further, a fixed amplitude is given in advance to each sound source position in the speech code based on the number of selectable candidates for each sound source position, and the sound sources arranged at the sound source positions are multiplied by this fixed amplitude, Since the driving sound source is generated by performing the above addition, the waste of the amplitude for each sound source is reduced with a simple configuration, and an effect that a high-quality speech decoding apparatus can be provided is obtained.

Embodiment 6 FIG.
FIG. 13 is a block diagram showing the configuration of the drive excitation encoding means 5 in the speech encoding apparatus according to Embodiment 5 of the present invention. The overall configuration of the speech encoding apparatus is the same as that shown in FIG. In FIG. 13, reference numeral 42 denotes pre-table correction means. In this embodiment, by adding only the pre-table correction means 42, the perceptually weighted encoding target signal is orthogonalized with respect to the adaptive sound source.

Next, the operation will be described.
First, a quantized linear prediction coefficient is input from the linear prediction coefficient coding means 3 in the speech coding apparatus to the auditory weighting filter coefficient calculation means 16 and the basic response generation means 18 in the driving excitation coding means 5. The The adaptive excitation repetition period obtained by converting the adaptive excitation code is input from the adaptive excitation encoding means 4 to the basic response generation means 18 in the driving excitation encoding means 5. In addition, the adaptive excitation encoding unit 4 inputs to the auditory weighting filter 17 in the driving excitation encoding unit 5 the input sound 1 or a signal obtained by subtracting the synthesized sound of the adaptive excitation from the input sound 1 as an encoding target signal. . The adaptive excitation is input from the adaptive excitation encoding unit 4 to the pre-table correction unit 42 in the driving excitation encoding unit 5.

  The perceptual weighting filter coefficient calculating means 16 calculates perceptual weighting filter coefficients using the quantized linear prediction coefficients, and sets the perceptual weighting filter coefficients as the perceptual weighting filter 17 and perceptual weighting filter 19 filter coefficients. . The perceptual weighting filter 17 performs a filtering process on the input encoding target signal with the filter coefficient set by the perceptual weighting filter coefficient calculation means 16.

  The basic response generation means 18 performs a periodic process using the repetition period of the input adaptive sound source on the unit impulse or the fixed waveform, and uses the obtained signal as a sound source to obtain the quantized linear prediction coefficient. A synthesized sound is generated by a synthesis filter configured by using it, and this is output as a basic response. The auditory weighting filter 19 performs a filtering process on the input basic response with the filter coefficient set by the auditory weighting filter coefficient calculation means 16.

  The pre-table calculation means 20 uses a signal with a predetermined sound source arranged at one sound source position as a temporary drive sound source, and correlates the perceptually weighted encoding target signal with the perceptually weighted basic response, that is, perceptually weighted. The correlation value of the synthesized sound based on the temporary driving sound source corresponding to all the sound source position candidates weighted to be encoded and the perceptually weighted sound source is calculated as d (x), and the cross correlation value of the perceptually weighted basic response, that is, Then, the cross-correlation value between the synthesized sounds based on the temporary drive sound source corresponding to all candidate combinations is calculated as φ (x, y). These d (x) and φ (x, y) are stored as a pre-table.

  The pre-table correction means 42 inputs the adaptive sound source and the pre-table stored in the pre-table calculation means 20, performs correction processing based on the following equations (12) and (13), and the obtained result is Thus, d ′ (x) and φ ′ (x, y) for each sound source position are obtained by the equations (14) and (15), and these are newly stored as a pre-table.

However,
c _tgt is a correlation value between the perceptually weighted encoding target signal and perceptually weighted adaptive sound source response (synthesized sound), that is, perceptually weighted encoding target signal and perceptually weighted synthetic sound based on the adaptive sound source. Correlation value between
c _x is a correlation value between a signal in which an auditory weighted basic response is arranged at the sound source position x and an auditory weighted adaptive sound source response (synthesized sound), that is, a synthesized sound based on temporary drive sound sources corresponding to all sound source position candidates. And the correlation value between the synthesized sound based on the adaptive sound source,
p _acb is the power of the adaptive sound source response (synthetic sound) weighted perceptually.

  Finally, the search means 21 sequentially reads out the sound source position candidates from the sound source position table 22, and calculates the evaluation value D for each combination of the sound source positions based on the expressions (1), (4), and (5). Calculation is performed using the pre-table stored by the pre-table correction means 42, that is, d ′ (x) and φ ′ (x, y) for each sound source position. Then, a combination of sound source positions that maximizes the evaluation value D is searched, and the obtained sound source position codes (indexes in the sound source position table) and polarities representing the plurality of sound source positions are output as drive sound source codes and this driving is performed. A time series vector corresponding to the excitation code is output as a driving excitation.

As described above, according to the sixth embodiment, the correlation value c _tgt between the encoding target signal and the synthesized sound based on the adaptive sound source, the synthesized sound based on the temporary driving sound source corresponding to all sound source position candidates, and Since the correlation value c _x with the synthesized sound based on the adaptive sound source is obtained and the pre-table is corrected using these values, the auditory weighted code is obtained without increasing the processing amount in the search means 21. The signal to be encoded can be orthogonalized with respect to the adaptive sound source, thereby improving the encoding characteristics and providing an effect of providing a high-quality speech encoding apparatus.

It is a block diagram which shows the structure of the drive excitation encoding means in the audio | voice coding apparatus by Embodiment 1 of this invention. It is a block diagram which shows the structure of the drive excitation decoding means in the audio | voice decoding apparatus by Embodiment 1 of this invention. It is a figure explaining the relationship between the encoding object signal by Embodiment 1 of this invention, and the sound source position of the periodic drive sound source. It is a figure explaining the relationship between the encoding object signal by Embodiment 1 of this invention, and the sound source position of the periodic drive sound source. It is a block diagram which shows the structure of the drive excitation encoding means in the audio | voice coding apparatus by Embodiment 2 of this invention. It is a block diagram which shows the structure of the drive excitation decoding means in the audio | voice decoding apparatus by Embodiment 2 of this invention. It is a figure explaining the adaptive sound source produced | generated by the adaptive sound source production | generation means by Embodiment 2 of this invention. It is a figure explaining the adaptive sound source produced | generated by the adaptive sound source production | generation means by Embodiment 2 of this invention. It is a figure explaining the adaptive sound source produced | generated by the adaptive sound source production | generation means by Embodiment 2 of this invention. It is a block diagram which shows the structure of the drive excitation encoding means and the perceptual weighting control means in the audio | voice coding apparatus by Embodiment 3 of this invention. It is a block diagram which shows the structure of the drive excitation encoding means and the perceptual weighting control means in the audio | voice coding apparatus by Embodiment 4 of this invention. It is a figure which shows the sound source position table by Embodiment 5 of this invention. It is a block diagram which shows the structure of the drive excitation encoding means in the audio | voice coding apparatus by Embodiment 6 of this invention. It is a block diagram which shows the structure of the conventional CELP type | system | group audio | voice coding apparatus. It is a block diagram which shows the structure of the conventional CELP type | system | group audio | voice decoding apparatus. It is a figure which shows the position candidate of the conventional pulse sound source. It is a block diagram which shows the structure of the drive excitation encoding means in the conventional CELP type | system | group audio | voice encoding apparatus. It is a figure explaining the relationship between the conventional encoding object signal and the sound source position of the periodic drive sound source. It is a figure explaining the relationship between the conventional encoding object signal and the sound source position of the periodic drive sound source.

Explanation of symbols

  1 input speech, 2 linear prediction analysis means, 3 linear prediction coefficient coding means, 4 adaptive excitation coding means, 5 driving excitation coding means, 6 gain coding means, 7 multiplexing means, 8 speech codes, 9 separation means 10 linear prediction coefficient decoding means, 11 adaptive excitation decoding means, 12 driving excitation decoding means, 13 gain decoding means, 14 synthesis filter, 15 output speech, 16 perceptual weighting filter coefficient calculation means, 17, 19 perceptual weighting Filter, 18 Basic response generation means, 20 Pre-table calculation means, 21 Search means, 22 Sound source position table, 23 Period preliminary selection means, 24 Constant table, 25 Comparison means, 26 Pre-selection means, 27 Drive excitation coding means, 28 Period encoding means, 29 period decoding means, 30 drive excitation decoding means, 31 period preliminary selection means, 2 constant table, 33 adaptive excitation codebook, 34 adaptive excitation generation means, 35 distance calculation means, 36 preliminary selection means, 37 auditory weighting control means, 38 comparison means, 39 intensity control means, 40 auditory weighting control means, 41 average value Update means, 42 Pre-table correction means.

Claims (2)

A speech code that encodes input speech frame by frame and outputs a speech code using an adaptive sound source generated from a past sound source and a driving sound source configured with a predetermined number of pulse positions and a polarity corresponding to the pulse positions. In the conversion device,
A first impulse response when an impulse is set at a first pulse position among a plurality of pulse positions, and a second impulse when an impulse is set at a second pulse position among the plurality of pulse positions Correlation value calculating means for calculating a first correlation value with a response for all combinations of the first pulse position and the second pulse position;
A second correlation value between an impulse response when an impulse is raised at one of the plurality of pulse positions and a synthesized sound based on the adaptive sound source is calculated, and the second correlation value is used. Correlation value correcting means for correcting the first correlation value;
Search means for calculating an evaluation value related to coding distortion using the corrected first correlation value and searching for the predetermined number of pulse positions of the driving sound source using the evaluation value;
The speech coding apparatus according to claim 1, wherein the correlation value correcting means performs correction using a product of the power of the synthesized sound based on the adaptive sound source and the second correlation value. A speech code that encodes input speech frame by frame and outputs a speech code using an adaptive sound source generated from a past sound source and a driving sound source configured with a predetermined number of pulse positions and a polarity corresponding to the pulse positions. In the process
A first impulse response when an impulse is set at a first pulse position among a plurality of pulse positions, and a second impulse when an impulse is set at a second pulse position among the plurality of pulse positions A correlation value calculating step of calculating a first correlation value with a response for all combinations of the first pulse position and the second pulse position;
A second correlation value between an impulse response when an impulse is raised at one of the plurality of pulse positions and a synthesized sound based on the adaptive sound source is calculated, and the second correlation value is used. A correlation value correcting step for correcting the first correlation value;
A search step of calculating an evaluation value related to encoding distortion using the corrected first correlation value, and searching for the predetermined number of pulse positions of the driving sound source using the evaluation value;
In the speech coding method, the correlation value correcting step performs correction using a product of the power of the synthesized sound based on the adaptive sound source and the second correlation value.

Images