JP2943983B1 - Audio signal encoding method and decoding method, program recording medium therefor, and codebook used therefor - Google Patents

Abstract

Abstract: ACELP (Algebric CEL)
P) to achieve a high compression ratio, reduce the amount of memory and computation,
And keep high quality. A positive / negative sign book i corresponding to a pulse position candidate of each position code book i (i = 1, 2,...) Is determined (for example, see FIG. 3). In the example of FIG. 3, the even-numbered sample points (position candidates) have a positive polarity, and the odd-numbered sample points have a negative polarity. When a pulse is generated at an adjacent sample point, the polarity of the pulse is opposite to each other. The position codebook i and the positive / negative codebook i are read with the same position code i.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to predictive coding for synthesizing speech by driving a filter representing a spectral envelope characteristic of an audio signal such as speech or music with a sound source vector, thereby reducing a signal sequence of the speech to a small amount of information. The present invention relates to a high-efficiency speech encoding method for digital encoding by using the method, a decoding method thereof, a codebook used for these methods, and a program recording medium therefor.

[0002]

2. Description of the Related Art In digital mobile communication, a high-efficiency voice encoding method is used in order to efficiently use radio waves or to efficiently use a communication line or a storage medium for a voice or music storage service. . At present, as a method for efficiently encoding speech, an original speech is divided into sections called frames or subframes at a constant interval of about 5 to 50 ms, and the speech of one frame is subjected to a linear filter representing envelope characteristics of a frequency spectrum. And a method of separating the information into two pieces of information, ie, a drive excitation signal for driving the filter, and encoding each of the information. In this method, as a method of encoding a drive excitation signal, a method of separating and encoding a periodic component considered to correspond to a pitch period (fundamental frequency) of a voice and other components is known. As an example of the coding method of the drive excitation information, code-driven linear prediction coding (Code-Excited L
inner Prediction (CELP). For details of the above technique, see Reference M. R. Sch
roeder and B.R. S. Atal, “Code
-Excited Linear Prediction
n (CELP): High Quality Speed
chat VeryLow Bit Rates ”,
IEEE Proc. ICASP-85, pp. 93
7-940, 1985.

FIG. 8 shows an example of a functional configuration of the encoding method. The speech input to the input terminal is output to the linear prediction analysis unit 1-
At 1, a linear prediction parameter representing the frequency spectrum envelope characteristic of the input speech is calculated. The obtained linear prediction parameters are encoded in the linear prediction parameter encoding unit 1-2 and sent to the linear prediction parameter decoding unit 1-3. When distortion calculation is performed using spectral information of input speech, for example, in consideration of auditory characteristics in distortion calculation, the linear prediction parameter is also sent to the distortion calculation unit 1-6. The linear prediction parameter decoding unit 1-3 reproduces a synthesis filter coefficient from the received code, and sends it to the synthesis filter 1-5. When considering auditory characteristics in distortion calculation,
Instead of using the linear prediction parameters before quantization in the distortion calculator 1-6, the decoded linear prediction parameters may be used for distortion calculation. The details of the linear prediction analysis and examples of encoding the linear prediction parameters are described in, for example, “Digital Speech Processing” by Sadahiro Furui (Tokai University Press). Here, the linear prediction analysis unit 1-1, the linear prediction parameter encoding unit 1-2, the linear prediction parameter decoding unit 1-3, and the synthesis filter 1-5 may be replaced with non-linear ones.

[0004] The drive excitation vector generation section 1-4 generates a drive excitation vector candidate having a length of one frame and sends it to the synthesis filter 1-5. FIG. 9 shows a functional configuration example of the driving sound source vector generation unit 1-4. From the adaptive codebook 2-1, the immediately preceding past drive excitation vector (the drive excitation vector for the immediately preceding one to several frames that has been already quantized) c (t−1) stored in the buffer is stored in a certain cycle. By cutting out the cut-out vector at a corresponding length and repeating the cut-out vector until the length of the frame is reached, a time-series vector candidate corresponding to the periodic component of the voice is output. As the “certain period”, a period that reduces the distortion d in the distortion calculator 1-6 is selected. In general, the selected period generally corresponds to the pitch period of voice. The fixed codebook 2-2 outputs a time-series code vector candidate having a length of one frame corresponding to the aperiodic component of the voice. For these candidates, a predetermined number of candidate vectors are stored in accordance with the number of bits for encoding independently of the input speech. The fixed code vector candidates output from the fixed codebook 2-2 are periodicized as necessary by the periodicization unit 2-8 at a period (generally equivalent to a pitch period as described above) specified by the periodic code. You. What is periodicity?
This means that a comb filter having taps is applied to a specified period position, or a vector cut out from the head of the vector at a length corresponding to the specified period is repeated, as in the adaptive codebook. The periodic unit 2-8 is often used from the viewpoint of improving coding efficiency, but may not be used. Further, when there is no or little pitch component in the voice itself, such as in a consonant section, the periodic unit may not perform any function. The time series vector candidates output from the adaptive codebook 2-1 and the periodization unit 2-8 are multiplied by the weights generated by the weight generation units 2-3 in the multiplication units 2-4 and 2-5, respectively. , And are added in the adder 2-6 to become a drive excitation vector candidate c. In the configuration example of FIG. 9, the configuration may be such that only the fixed codebook 2-2 is used without using the adaptive codebook 2-1. When encoding a signal with a small pitch periodicity such as a consonant part or background noise, In order to save bits, a configuration not using the adaptive codebook 2-1 is often used.

The synthesis filter 1-5 in FIG. 8 is a linear filter using the output of the linear prediction parameter decoding section 1-3 as a filter coefficient, and outputs a reproduced sound candidate y with a driving excitation vector candidate c as an input. . Generally, the order of the synthesis filter 1-5, that is, the order of the linear prediction analysis, is generally about 10 to 16 order. As described above, the synthesis filter 1-5 may be a non-linear filter.

[0006] In the distortion calculation unit 1-6, the synthesis filter 1-
Then, a distortion d between the reproduced voice candidate y, which is the output of No. 5, and the input voice x is calculated. The calculation of this distortion is often performed taking into account the coefficients of the synthesis filter or unquantized linear prediction coefficients, for example, perceptual weighting. FIG.
Fig. 7 shows an example of a functional configuration for calculating distortion in consideration of auditory weighting. The perceptual weighting is configured in the form of perceptual weight filters 4-2 and 4-3 using unquantized linear prediction parameters or quantized synthetic filter coefficients. The reproduced voice candidate y output from the synthesis filter 4-1 is passed through the auditory weight filter 4-2, and the distortion d between the input voice and the input voice also passed through the auditory weight filter 4-3.
Is calculated. Here, the auditory weight filters 4-2, 4-
3 normally use the same filter coefficient, the auditory weighting filters 4-2 and 4-3 are equivalent even if they are inserted as one filter after the distance calculation unit 4-4. As shown in FIG. 10, it is often divided into two places before the distance calculation unit 4-4.

The codebook search control section 1-8 in FIG. 8 selects a driving excitation code that minimizes the distortion d between each reproduced speech candidate y and the input speech x, and determines the driving excitation vector in the frame. decide. When the adaptive codebook 2-1, the fixed codebook 2-2, and the weighted codebook 2-3 shown in FIG. 9 are used, a periodic code, a fixed code, and a weighted code for these are selected and driven. It is a sound source code.

The excitation code (periodic code, fixed (noise) code, weight code) determined by the codebook search control section 1-8 and the linear prediction parameter code output from the linear prediction parameter coding section 1-2. Is sent to the code sending section 1-9 and stored in a storage device or sent to the receiving side via a communication path according to the form of use. FIG. 11 shows a functional configuration example of a decoding method corresponding to the encoding method. Among the codes received from the transmission path or the storage medium, the linear prediction parameter code is decoded into a synthesis filter coefficient in the linear prediction parameter decoding unit 3-2, and the synthesis filter coefficient is decoded.
4 and, if necessary, to the post-processing unit 3-5. The received excitation code is used as a driving excitation vector generation unit 3-
3 and an excitation vector corresponding to the code is generated. The configuration of the driving excitation vector generation unit 3-3 is the same as that of the driving excitation vector generation unit 1-4 of the encoding method shown in FIG.
Is a configuration corresponding to. The synthesis filter 3-4 reproduces a sound by using the driving sound source vector as an input. Post-processing unit 3
-5 performs processing (also referred to as post-filtering) to aurally reduce the sense of noise in the reproduced sound, but the post-processing unit 3-5 may not be used due to processing amount reduction or the like. Many.

[0009]

Problems that arise in the CELP method are that a great deal of arithmetic processing is required for distortion calculation for selecting a driving excitation vector candidate, and that the fixed codebook 2- is used. 2 in that a very large amount of memory is required to store the code vector stored in the second.

[0010] To solve this problem, Algebraic
Code-Excited Linear Predi
ction (ACELP) has been proposed. In this method, instead of storing a fixed codebook as a vector pattern of a frame length, several pulses having a height of 1 are provided in a frame, for example, four pulses for a frame or a subframe of 80 samples, By adopting this driving excitation method in a method of setting a fixed code vector by setting it in an appropriate position, and devising the calculation order in distortion calculation, the amount of calculation processing and memory required is reduced compared to the conventional method. be able to. In addition, ACEL
For details of the P method, see, for example, Literature, Salami,
C. Laflame, and JP. Adoul,
“8 kbit / s ACELP Coding of
Speech with 10ms Speech-F
name: a Candidate for CCIT
T Standardization ”, IEEE P
rc. ICASP-94, pp. II-97.

Now, a specific example of ACELP will be described with reference to the drawings. FIG. 12 is a functional configuration example of a fixed codebook in ACELP. When N pulses are generated, N position codebooks 1 to N (5-11 to 5-1N), N pulse generators 1 to N (5-21 to 5-2N), and N positive and negative signs It has sign multiplication units 1 to N (5-31 to 5-3N). The position codebook 5-1 holds a table of position candidates where a pulse can be set for each pulse generation unit 5-2, and associates position codes with actual pulse positions. Pulse generator 5-
In step 2, a time-series vector for one subframe (or one frame) in which a single pulse is set at a sample point specified by the position codebook 5-1 is output. The sign multiplication unit 5-3 multiplies the pulse output from the pulse generation unit 5-2 by a sign of +1 or -1. The pulses multiplied by this code are added to form a vector composed of N pulses.

FIG. 13 shows an example in which the subframe length is 80.
This is to make it possible to schematically understand how the position codebook and the pulse generator generate pulses when the number of samples and pulses is four, so that two pulses do not overlap at the same sample point. This is an example of designing a pulse position codebook (position candidate table). For example, pulse 1 is
.. 75 are set at any one of the candidate positions (sample points). Similarly, pulses 2, 3,
4 is one of the position candidates (one of the broken lines)
Book). Each pulse is supplied to a plus / minus sign multiplier 5-3.
, The sign of +1 or -1 is multiplied, so that there is a case where a pulse is made in the plus direction at the same position and a case where a pulse is made in the minus direction. Thus,
A pulse train of four pulses with positive and negative signs is output. The distortion calculator 1-6 and the codebook search controller 1- in FIG.
In step 8, it is determined which pulse has the lowest distortion at which position in these pulse trains and the optimum position and sign are determined. The position code and sign are determined by the communication path or the storage medium. Through to the decoder. If the number of pulses is four or less in the example of FIG. 13, for example, there is a choice that the third pulse is not raised, and the number of pulses may be three or less.

The problem of the ACELP system is that, since the model of the fixed code vector is simple, if the number of pulses per fixed time length is reduced, that is, if the bit rate is reduced, the sound quality is significantly deteriorated. Is a point. For example, in the example of FIG.
4 is 4 bits each, pulse 3 is 5 bits, and 4 pulses require 4 bits to represent a positive / negative sign, for a total of 21 bits. If the number of pulses is reduced from four to three in order to reduce the number of bits, the required number of bits becomes 17 bits in total, but the sound quality is significantly degraded.

On the other hand, the conventional CELP has the advantage that the shape of the code vector stored in the codebook can be analyzed in advance by analyzing a large amount of speech data, and the vector having an effective shape can be stored efficiently. is there. In other words, it is not necessary to represent a shape that appears infrequently in real speech as a code vector, so that the degree of freedom of the vector shape can be constrained to reflect the properties of real speech, so that an efficient code with few bits can be used. Can be realized. But A
In CELP, since only pulses having a height of 1 are arranged, a statistical tendency included in actual speech cannot be reflected in codebook design. Therefore, it is necessary to choose between using a conventional CELP to increase the compression ratio or using an ACELP type method to reduce the amount of computation and memory.

SUMMARY OF THE INVENTION An object of the present invention is to provide a high compression rate (low bit rate) and inexpensive processor capable of obtaining high quality reproduced voice with a small amount of memory and a small amount of computation within an allowable range. It is an object of the present invention to provide a method of digitally encoding an audio signal such as voice or music, a decoding method thereof, these devices, and a program recording medium thereof.

[0016]

According to the present invention, ACEL is used.
In the P-type pulse-type coding method, the arrangement of the pulses is constrained so as to reflect the characteristics of the actual voice. That is, each position codebook is associated with a positive / negative codebook, and each pulse in the position codebook is assigned. Positive or negative polarity is given to the position candidate in advance, the sign code is also read out by the position code, and when the pulses stand side by side at adjacent sample points, they are made to have opposite polarities and appear in the actual voice. Infrequent patterns should not be generated. As a result, the conventional AC
Compared to the ELP type coding method, the bit rate can be reduced when the same quality is realized, and high quality coding can be realized when the bit rates are the same.

[0017]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 shows an example of a functional configuration of a fixed codebook according to the present invention. The difference from the conventional method shown in FIG. 12 is that the sign of each pulse is predetermined in accordance with the position of the pulse, and the conventional method specifies the sign of each pulse using one bit per pulse. On the other hand, in the present invention, the position and the positive / negative direction of the pulse are determined as a set only by the position code of the pulse, that is, the position codebooks 5-11 to 5-1N are changed to the positive and negative codebooks 7-11 to 7-7. The plus / minus sign for each pulse position which can be respectively specified with, for example, the position codebook 5-11 and which is combined with -1N is predetermined in the plus / minus codebook 7-11. Each of the output pulses of the pulse generation units 5-21 to 5-2N in the positive / negative sign multiplying units 5-31 to 5-3N is multiplied by the positive / negative sign from the positive / negative sign book 7-11 to 7-1N, respectively. If necessary, the N pulses determined in this way are combined, and the entire sign multiplying unit 7-2 multiplies the sign by one sign.

As a result, the number of bits required when using this codebook configuration example is N-1 bits smaller when using N pulses than when using the conventional method shown in FIG. The configuration example of FIG. 1 is equivalent even if modified as shown in FIG. 2. Finally, instead of multiplying the entire pulse train by plus and minus signs, the sign inverting units 1 to N
In (8-11 to 8-1N), the sign books 1 to N
If the sign of each sign from (7-11 to 7-1N) is negative, the polarity of the pulse is inverted, and if the sign is positive, the pulse is not inverted and processed as it is, and the sign multipliers 1 to N (5
-31 to 5-3N).

When the configuration examples shown in FIGS. 1 and 2 are used, the sign book 7-11 to 7-1N and the position code book 5
It is important how to determine the set of the position and the polarity of the pulse stored in -11 to 5-1N. Regardless of the method of determination, it is the same that N-1 bits can be reduced. However, when speech is encoded and decoded, it is necessary to determine the quality as little as possible.
As described above, in order to minimize the deterioration of the sound quality, when encoding of real voice is performed, the combinations that appear frequently are left, and the combinations that appear less frequently are reduced. A reduction method that does not reflect the properties of the actual voice will significantly degrade the voice quality.

FIG. 3 shows an example of a set of pulse positions and pulse polarities when the present invention is applied to the same position (sample point) that can be taken in the same manner as in the prior art shown in FIG. In this case, the required number of bits is 18 bits.
3 bits less than 3 bits. The position of the upward dashed line indicates that the pulse polarity is specified only in the plus direction, and the position of the downward dashed line indicates that the pulse polarity is specified only in the negative direction. However, since the last four pulses are multiplied together by plus and minus signs, this polarity is not an absolute sign but only a relative polarity between the four pulses. In the example of FIG. 3, the pulse 1 is specified to have a positive polarity in the even sample position and a negative polarity in the odd sample position. Even-numbered sample points may have negative polarity, and odd-numbered sample points may have positive polarity.

At first glance, the designation as shown in FIG. 3 does not seem to have contributed to not deteriorating the sound quality.
This is an effective specification method that accurately reflects the characteristics of actual speech. When observing the shape of the driving sound source vector when actual speech is coded, pulses of opposite polarities often occur at adjacent sample points, and pulses of the same polarity rarely line up at adjacent sample points. Absent. The designation method shown in FIG. 3 accurately reflects this fact. According to this designation method, when pulses stand side by side at adjacent sample points, the polarities of the two pulses are always opposite. Pulses of the same polarity do not stand adjacent to each other. Therefore, although it is an algebraic method of determining even-numbered sample points and odd-numbered sample points, the method well reflects the characteristics of speech.

FIG. 4 shows a candidate example 2 as a modification of the candidate example of FIG. The difference from the example of FIG. 3 is that at sample points 15, 31, 47, and 63 indicated by crosses, the pulse amplitude is 0,
That is, no pulse is generated. This is because the quality when encoding is better when three or less options are prepared, rather than always setting four pulses in one subframe. The determination as to whether four pulses or three pulses or less is preferable is made by the distortion calculator 1-6 and the codebook search controller 1
-8 determines the smaller distortion. As shown in FIG. 4, when a sample point where a pulse does not occur is formed, the quality of the encoded sound is better when the position where the pulse does not occur is dispersed in the subframe, and the sample point 1
The determination method of 5, 31, 47, 63 is a typical example of a good determination method.

FIG. 5 shows an example of position and polarity candidates when it is desired to further reduce one bit per subframe to 17 bits. No pulse is raised at sample positions without sample point numbers in the figure, that is, 4, 9, 14, 19,..., 74, 79. In addition, no pulse is made on the crosses 31 and 47. Although the quality of the coded sound is inferior to that of the configuration example of FIG. 4, it is a good example from the viewpoint that it is better to disperse the sample points where a pulse cannot be formed in a subframe. FIG. 4 shows the case of 18 bits, and FIG. 5 shows the case of 17 bits. For example, the first subframe uses the configuration of FIG. 5 by allocating 17 bits to a fixed codebook, and the second subframe has 18 bits. 4, different configurations may be mixed in one encoding, such as using the configuration of FIG.

FIG. 6 shows a modification of FIG. Figures 3-5
Is characterized in that the relative polarity of the pulse is determined depending on whether the sample point is an even or odd point, but in FIG.
The relative polarity of the pulse is not determined by the even sample point or the odd sample point. However, as in FIGS. 3 to 5, when the pulses stand side by side at adjacent sample points, the polarities of the two pulses are always opposite, and pulses of the same polarity do not stand adjacent. Therefore, the method of setting the pulse shown in FIG. 6 is also an effective designation method that accurately reflects the characteristics of the actual voice. A further advantage of this method is that pulse 1 has the same polarity at all the candidate positions, and that pulses 2 to 4 have the same polarity at all the respective candidate positions. There is a feature that the search processing of the pulse position is simplified. In FIG. 6, a sample position at which no pulse is generated may be provided as shown by a cross in FIG.

FIG. 7 shows another example of the functional configuration of the fixed codebook according to the present invention.
Or −1 pulse, but the sign books 1 to N (7-1)
1 to 7-1N), instead of amplitude codebooks 1 to N (9-1)
1 to 9-1N), the positive and negative polarities of the pulse and the pulse height are determined in advance for each pulse position specified by the position code. The amplitude codebooks 9-11 to 9-1N output pulse amplitude values at corresponding sample points from the position code. For example, when a pulse rises at the sample point 10, the pulse height is +0.6,
When a pulse rises at the sample point 57, the pulse height is -0.2. In the configuration example of FIG. 2, the amplitude codebooks 9-11 to 9-1N in FIG.
Is equivalent to having only the value of. Amplitude codebook 1-N
The design of (9-11 to 9-1N) needs to be designed so that the quality of the coded sound is as good as possible. In general, a pattern having a high appearance frequency when an actual voice is coded is prepared. Is good. It is good to observe the driving sound source when using actual audio data, such as displaying a waveform, and prepare a pattern with a high frequency of appearance.However, using a method called the generalized Lloyd algorithm, using a large amount of audio data , A method of minimizing the sum of distortions can be applied. The generalized Lloyd algorithm is a method that starts from a certain initial amplitude codebook and updates the amplitude codebook so that the sum of distortions when encoding all audio data is minimized. Are alternately repeated to form an optimum codebook gradually. When the generalized Lloyd algorithm is applied to the present invention, the initial codebook is determined by the position and amplitude (+
It is a good representative example to set 1 or -1) as the initial value.

In FIG. 7, sign inverting sections 8-11 to 8-1 are provided.
N is omitted, and as shown in FIG.
The respective outputs of 1 to 9-1N may be multiplied by multipliers 5-31 to 5-3N, respectively, added, and then multiplied by the plus / minus sign by the multiplier 7-2. The above-described encoding processing and decoding processing can be performed by reading a program by a computer and executing the decoding.

[0027]

In order to confirm the effects of the present invention,
A computer program for encoding and decoding actual audio was created. The present invention may be realized by hardware or in the form of a computer program, but this time, a computer program was used as a method for easily confirming the effects. The sampling frequency is 8
kHz, subframe length is 10 ms (80 samples),
One frame has a two-subframe configuration, the codebook of FIG. 5 is used for the first subframe, the codebook of FIG. 4 is used for the second subframe, and the bit rate is 4 kbit / s. The actual sound data was input to the above program, encoded and decoded, and the quality of the reproduced sound was examined.
TU-T G. The quality almost equal to the quality in the case of using the 726 standard system could be realized. G. FIG. Since the bit rate of the 726 standard system is 32 kbit / s, equivalent quality can be realized at a bit rate of 1/8. The CPU time required for encoding is the same as the conventional CELP
It has been confirmed that high-quality speech encoding can be realized with less than 1/4 of the system (more than 4 times as fast) and with a small amount of processing.

[Brief description of the drawings]

FIG. 1 is a diagram showing a functional configuration example of a fixed codebook according to the present invention.

FIG. 2 is a diagram showing a modification of the functional configuration of the fixed codebook according to the present invention.

FIG. 3 is a diagram showing pulse position and polarity candidate examples according to the present invention.

FIG. 4 is a diagram showing another example of pulse position and polarity candidates according to the present invention.

FIG. 5 is a diagram showing still another example of pulse position and polarity candidates according to the present invention.

FIG. 6 is a diagram showing still another example of pulse position and polarity candidates according to the present invention.

FIG. 7 is a diagram showing still another example of the functional configuration of the fixed codebook according to the present invention.

FIG. 8 is a diagram showing a functional configuration example of a conventional speech coding method.

FIG. 9 is a diagram showing a functional configuration example of a driving sound source vector generation unit 1-4 in FIG. 8;

FIG. 10 is a diagram showing an example of a functional configuration of a combined distortion calculator 1-7 in FIG. 8;

FIG. 11 is a diagram showing a functional configuration example of a CELP-type speech decoding method.

FIG. 12 is a diagram showing an example of a functional configuration of a fixed codebook of ACELP.

FIG. 13 is a diagram showing a conventional example of pulse position candidates.

──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Shoko Kurihara 3-19-2 Nishishinjuku, Shinjuku-ku, Tokyo Within Nippon Telegraph and Telephone Corporation (56) References JP-A-2-282799 (JP, A) JP-A-3-177900 (JP, A) JP-A-3-245197 (JP, A) JP-A-8-54898 (JP, A) JP-A-10-97294 (JP, A) (58) Int.Cl. ⁶ , DB name) G10L 9/14, 9/18 H03M 7/30

Claims (13)

(57) [Claims] An audio signal is reproduced by driving a synthesis filter with a driving excitation vector obtained from a time series vector extracted from a codebook, and a quantization unit of an excitation vector called a frame or a subframe is used as a driving excitation vector. Generate two or more pulses in the section, and search for the position, positive / negative polarity, and amplitude of the pulse so that the distortion between the input audio signal and the reproduced audio signal is minimized or conforms to the minimum. In the coding method to be determined, the relative polarity of a plurality of pulses to be set in a frame or subframe is determined in advance depending on the sample position where the pulse is to be set, and the relative polarity of the pulse is determined at the same time when the pulse position is determined. , Which is also uniquely determined. 2. The method for encoding an audio signal according to claim 1, wherein the position of each pulse in the frame or the sub-frame and the relative polarity of each pulse are determined, and all the pulses in the frame or the sub-frame are determined. A method for encoding an acoustic signal, comprising: determining an absolute polarity of each pulse by designating one kind of positive / negative sign collectively. 3. The audio signal encoding method according to claim 1, wherein a positive or negative sign depending on a pulse position is set so that pulses having the same polarity do not stand side by side at adjacent sample points. A method for encoding an audio signal, comprising using a codebook. 4. The encoding method of an acoustic signal according to claim 1, wherein the polarity of the pulse is determined depending on whether the sample position where the pulse is to be formed is an even sample point or an odd sample point. ,
A coding method for an acoustic signal, comprising using a codebook of a positive / negative sign depending on a pulse position so that one is positive and the other is negative. 5. The method for encoding an acoustic signal according to claim 1, wherein the absolute value of the pulse amplitude is 1. 6. The method of encoding an acoustic signal according to claim 1, wherein real values of the polarity and amplitude of the pulse at each pulse position are determined in advance and stored in a codebook. Wherein the polarity and amplitude of the pulse are uniquely determined simultaneously at the same time. 7. A linear prediction parameter code and a driving excitation code are input, the linear prediction parameter code is decoded into a synthesis filter coefficient, the synthesis filter coefficient is set in a synthesis filter, and the driving excitation code is set to a driving excitation vector. In the decoding method of reproducing the acoustic signal by driving the synthesis filter with the driving excitation vector, a position code is decoded from the driving excitation code, and a position codebook and a positive / negative codebook are decoded by the position code. A sound signal decoding method comprising: generating a drive excitation vector by generating a pulse having a unique polarity at a position corresponding to the position code. 8. A first to N-th position codebook which has a plurality of sample points different from each other as pulse position candidates and outputs pulse position candidates in accordance with respective given position codes; It is provided corresponding to the codebook, holds a predetermined positive / negative polarity for each pulse position candidate of the corresponding position codebook, and outputs positive or negative polarity according to the given position code. First to N-th codebooks, and pulse position candidates output from the first to N-th position codebooks, respectively, and first to N-th pulse generation for generating a pulse at the position of the pulse position candidate Means for multiplying each pulse output from the first to Nth pulse generation means by positive or negative polarity output from the first to Nth positive / negative codebook, respectively.
An audio signal encoding / decoding codebook comprising: a positive / negative sign multiplying means; and an adding means for combining all of the multiplied outputs of the first to Nth positive / negative sign multipliers. 9. The audio signal encoding / coding apparatus according to claim 8, further comprising: all-code multiplying means for multiplying an output pulse of said adding means by a given positive or negative sign. Codebook for decoding. 10. A first to Nth position codebook which has a plurality of sample points different from each other as pulse position candidates and outputs pulse position candidates corresponding to given position codes, respectively, and said first to Nth position codebooks. The code book is provided in correspondence with each other, and holds a positive or negative polarity for a predetermined pulse for each pulse position candidate of the corresponding position code book, and sets a positive or negative polarity according to the given position code. The first to Nth sign books to be output and the output positive or negative polarity of the first to Nth sign books are output with or without performing sign inversion according to a given common sign. First to N-th sign inverting means; and first to N-th pulses each receiving a pulse position candidate output from the first to N-th position codebooks and generating a pulse at the position of the pulse position candidate. Generating means; and first to N-th positive / negative signs for multiplying each of the pulses output from the first to N-th pulse generating means by positive or negative polarity output from the first to N-th sign inverting means, respectively. An audio signal encoding / decoding codebook, comprising: a multiplying unit; and an adding unit that combines all of the multiplied outputs of the first to N-th sign multiplying units. 11. A first to N-th position codebook which has a plurality of sample points different from each other as pulse position candidates and outputs position candidates corresponding to given position codes, respectively, and said first to N-th position codes. Each of the pulse position candidates of the corresponding position codebook is provided with an amplitude value including predetermined positive and negative polarities, and includes the polarity according to the given position code. First to Nth amplitude codebooks for outputting amplitude values, and respective pulse position candidates output from the first to Nth position codebooks, respectively, are inputted, and a first pulse for generating a pulse at the position of the pulse position candidate is input. To the N-th pulse generation means, and for the pulses output from the first to N-th pulse generation means, an amplitude value including the polarity from the first to N-th amplitude codebooks and a common positive / negative sign for the whole. It Means for multiplying each and synthesizing all the multiplication results and outputting the result. 12. A process for obtaining a linear prediction parameter representing a frequency spectrum envelope characteristic of an input audio signal; a process for encoding the linear prediction parameter to obtain a linear prediction parameter code; and decoding and combining the linear prediction parameter code. A process of obtaining a filter coefficient and setting the same in a synthesis filter; a process of reading out a position codebook and a positive / negative codebook by a position code to generate a plurality of drive excitation pulses; A process of synthesizing signals, a process of obtaining distortion of the synthesized audio signal with respect to the input audio signal, a process of searching for the position code so as to minimize the distortion and determining a driving excitation code, Outputting a linear prediction parameter code and the driving excitation code as encoded output by a computer. A recording medium recording a program to. 13. A process of decoding an input linear prediction parameter code to obtain a synthesis filter coefficient and setting the synthesis filter coefficient in a synthesis filter; a process of decoding a driving excitation code to obtain a position code; A program for performing, by a computer, a process of reading a position codebook and a positive / negative codebook with a position code to generate a plurality of drive excitation pulses, and a process of driving the synthesis filter by the drive excitation pulses to synthesize an acoustic signal. And a recording medium on which is recorded.