JP2008046405A - Encoding method and decoding method in adaptive differential pulse code modulation system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To further compress data while suppressing degradation of sound quality at the reproduction of voice signal. <P>SOLUTION: In an encoding method by an adaptive differential pulse code modulation (ADPCM) encoder, when it desires to determine that sampling condition is satisfied, based on α which is set to be a number where the sound quality will not be affected adversely during reproduction of voice by a comparator unit 29; since an ADPCM value Dn which is output from a coding unit 22 is thinned and this is not stored in a memory 30, the compressed data amount which is stored in the memory 30 can be reduced. Furthermore, in a decoding method by an ADPCM decoder, when it is determined that interpolation condition is satisfied, based on α which is set to a number so that the sound quality will not be degraded during the reproduction of voice by a comparator unit 32, since the value which is thinned by a comparator unit 29 is interpolated by an encoding unit 33, an original voice signal is reproduced accurately. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an encoding method and a decoding method of an adaptive differential pulse code modulation (hereinafter referred to as “ADPCM”) method, and more particularly to a data compression method thereof.

  The ADPCM method is a data compression method in which the difference between the previous (previous) sampling value and the current sampling value is encoded into a value of several bits (ADPCM value). At this time, the basic quantization width Δn is adaptively changed for each sampling, and the ADPCM value at that time is used to determine how many times the quantization width Δn is to be increased. Is improved as compared with the Delta Modulation (DM) method and the like.

FIG. 2 is a waveform diagram showing the operation principle of the conventional ADPCM method.
In FIG. 2, if the original waveform value (PCM) at the nth sampling value is Xn and the playback value of the previous n-1th ADPCM data is Yn-1, the difference value dn is
dn = Xn−Yn-1
It is represented by This is the quantization width Δn
Δn = Δn-1 ・ M (Dn-1)
Quantize with to find ADPCM value Dn. Here, M is a function of Dn−1 and is determined based on the statistical properties of the speech signal (speech waveform). For example, if the difference value dn is represented by a 4-bit ADPCM value D3D2D1D0 with the most significant bit (MSB) as a sign bit, the nth reproduction value Xn is represented as follows.
Xn = Yn-1 + qn (1)
qn = (-1) ^D3・ (D2 ・ Δn + 1/2 ・ D1 ・ Δn + 1/4 ・ D0 ・ Δn + 1/8 ・ Δn)
(2)
By repeating this sequentially, the ADPCM reproduction value at each n can be obtained.

FIG. 3 is a schematic configuration diagram showing a conventional ADPCM encoder.
In this ADPCM encoder encoding method, an input analog audio signal (audio waveform) is sampled by an analog / digital converter (hereinafter referred to as “A / D converter”) 1 to obtain a digital signal (nth element). Converted to waveform value Xn). The subtractor 2 subtracts the previous (n−1) th reproduced value Yn−1 from the value Xn of the original waveform to obtain a difference value dn. The difference value dn is adaptively quantized by the encoder (ENC) 3 based on the quantization width Δn calculated by the adaptation unit 4 to obtain the ADPCM value Dn.

The ADPCM value Dn is inversely quantized by the decoder (DEC) 5 to obtain a change amount qn, and this change amount qn is added to the previous reproduction value Yn-1 by the adder 6 to obtain the current reproduction. A value Yn is determined. A delay unit (Z- ¹ ) 7 obtains a reproduction value Yn-1 immediately before the current reproduction value Yn, and this reproduction value Yn-1 is fed back (feedback) to the adder 6 and the difference unit 2. The The ADPCM value Dn output from the encoder 3 is stored in the memory (MEM) 8.

FIG. 4 is a schematic configuration diagram showing a conventional ADPCM decoder.
In this decoding method of the ADPCM decoder, the ADPCM value Dn stored in the memory (MEM) 8 is adaptively dequantized by the decoder (DEC) 10 based on the quantization width Δn calculated by the adaptation unit 9. The current change amount qn is obtained. The current change amount qn and the previous reproduction value Yn-1 are added by the adder 11 to obtain the current reproduction value Yn. A delay unit (Z- ¹ ) 12 obtains a past reproduction value Yn-1 from the current reproduction value Yn and feeds it back to the adder 11. The current reproduction value Yn output from the adder 11 is converted into an analog audio signal by a digital / analog converter (hereinafter referred to as “D / A converter”) 13 to reproduce the original audio signal.

In such an ADPCM encoding method and decoding method, the function M of the quantization width Δn is a constant A and an integer Δpn,
M (Dn-1) = A ^{△ Pn-1}
, The quantization width Δn is
Δn = Δn-1 ・ A ^{△ Pn-1} = Δn-1Δn-2 ・ A ^{△ Pn-2} = ・ ・ ・
= △ 0 ・ A ^{△ Pn-1} (Pn-1 = Σ △ pi)
It becomes. Therefore, the above-described calculation can be reduced by making the quantization width Δn and Pn (pointer value indicating where to extract data from the memory) correspond one-to-one. In general, this is accomplished using a read-only memory (hereinafter “ROM”) table.

  A data compression method in such an ADPCM encoding method and decoding method is described in, for example, the following patent documents.

JP-A-5-335968 Japanese Patent Laid-Open No. 11-7300 JP 2001-175283 A

  These Patent Documents 1 to 3 describe a technique for compressing data by thinning data in the ADPCM method. That is, in Patent Document 1, thinning is performed by changing the sampling frequency, and in Patent Document 2, a predetermined number of samples in a sound part is blocked, and several samples before and after that are thinned out. Further, in Patent Document 3, data is thinned out at the time of fast-forwarding, and this is also thinned out for each voice block.

  However, in the conventional ADPCM encoding and decoding methods shown in FIG. 2 to FIG. 4, even if the change in the original waveform is small from the previous ADPCM reproduction value and can be approximated by interpolation, etc. Since the data is encoded into the ADPCM value Dn, the data cannot be further compressed. This is also true for the techniques of Patent Documents 1 to 3, and rough thinning is performed, and the data compression rate cannot be increased.

  As described above, in the conventional method, it is difficult to further compress the data of the audio signal by suppressing the deterioration of the sound quality during reproduction.

  In the ADPCM encoding method according to the first aspect of the present invention, the difference value dn between the original waveform value Xn of the digitized audio signal and the reproduction value Yn-1 of the previous reproduction waveform is obtained. The ADPCM value Dn and the pointer value Pn are obtained by quantizing the difference value dn with a quantization width Δn determined by the previous pointer value Pn−1, and the pointer value Pn−1 immediately before the pointer value Pn is obtained. And the next pointer value Pn + 1 is obtained. Then, based on the pointer values Pn-1, Pn, and Pn + 1, when the change in the quantization width Δn is smaller than a predetermined value compared to the preceding and subsequent data, the ADPCM value Dn is thinned out, and is larger than the predetermined value Stores the ADPCM value Dn as it is in the memory.

  In the ADPCM decoding method of the first invention, the quantization width Δn determined by the pointer value Pn is obtained from the ADPCM value Dn stored in the memory of the first invention, and the pointer immediately preceding the pointer value Pn is obtained. Find the value Pn-1. When the pointer values Pn−1 and Pn are smaller than a predetermined value, the ADPCM value Dn is dequantized and interpolated with a quantization width smaller than the quantization width Δn.

  In the ADPCM encoding method of the second invention, a difference value dn between the original waveform value Xn of the digitized audio signal and the reproduction value Yn-1 of the previous reproduction waveform is obtained, and the pointer value Pn is used. The ADPCM value Dn is obtained by quantizing the difference value dn with a predetermined quantization width Δn. When the difference value dn is smaller than the predetermined value, the ADPCM value Dn is thinned out. When the difference value dn is larger than the predetermined value, the ADPCM value Dn is stored in the memory as it is.

  In the ADPCM decoding method of the second invention, the quantization width Δn is obtained from the ADPCM value Dn stored in the memory of the second invention, and when the ADPCM value Dn is thinned, the quantum The ADPCM value Dn is inversely quantized and interpolated with a quantization width smaller than the quantization width Δn.

  According to the first invention, when ADPCM encoding is performed, the ADPCM data stored in the memory is thinned out by the pointer values Pn−1, Pn, and Pn + 1. Therefore, the compressed data stored in the memory The amount can be reduced. Furthermore, when ADPCM is decoded, ADPCM value Dn is inversely quantized with a small quantization width and interpolated based on pointer values Pn-1 and Pn, so that the original audio signal can be accurately reproduced. Accordingly, it is possible to increase the data compression amount by suppressing the deterioration of the sound quality at the time of reproducing the sound, compared with the conventional case.

  According to the second invention, when ADPCM encoding is performed, ADPCM data is thinned out based on a difference value from the previous ADPCM reproduction value, so that the amount of compressed data stored in the memory is reduced. I can do it. Further, when ADPCM is decoded, the quantization width Δn is obtained from the ADPCM value Dn stored in the memory. When the ADPCM value Dn is thinned out, the ADPCM value Dn is inversely quantized with a small quantization width. Therefore, the original audio signal can be reproduced with high accuracy. Accordingly, almost the same as in the first aspect of the present invention, it is possible to increase the data compression amount by suppressing the deterioration of the sound quality at the time of reproducing the sound as compared with the prior art.

  In the ADPCM encoding method, a difference value dn between the original waveform value Xn of the digitized audio signal and the reproduction value Yn-1 of the previous reproduction waveform is obtained, and the previous pointer value Pn-1 is obtained. The ADPCM value Dn and the pointer value Pn are obtained by quantizing the difference value dn with a quantization width Δn determined by: Further, the pointer value Pn−1 immediately before the pointer value Pn and the pointer value Pn + immediately after the pointer value Pn Ask for one. Then, based on the pointer values Pn-1, Pn, and Pn + 1, when the change in the quantization width Δn is smaller than a predetermined value compared to the preceding and subsequent data, the ADPCM value Dn is thinned out, and is larger than the predetermined value Stores the ADPCM value Dn as it is in the memory.

  Here, when the predetermined value in the range that the pointer value Pn can take is α, when α> Pn−1> Pn> Pn + 1, the ADPCM value Dn is thinned out.

(Configuration of Example 1)
FIGS. 1A and 1B are diagrams showing a data compression method in the ADPCM encoding method and decoding method according to the first embodiment of the present invention. FIG. 1A is a schematic diagram of an ADPCM encoder. The configuration diagram and FIG. 5B are schematic configuration diagrams of the ADPCM decoder.

The ADPCM encoder in FIG. 1A includes an A / D converter 20 that converts an input analog audio signal (audio waveform) into a digital signal (current nth original waveform value). The A / D converter 20 is connected to an encoder (ENC) 22 via a subtractor 21, and an adaptation unit 23 is feedback connected to the encoder 22. An adder 25 is connected to the encoder 22 via a decoder (DEC) 24, and a delay unit (Z- ¹ ) 26 is feedback-connected to the adder 25. The delay unit 26 is feedback-connected to the difference unit 21.

  In the first embodiment, a differencer 27, an encoder (ENC) 28, and a comparator (CMP) 29 are newly added on the output side of the A / D converter 20 and the adder 25. The comparator 29 compares the pointer values Pn,... Output from the encoders 22 and 28 with a predetermined value (for example, the number α) and thins out data. (MEM) 30 is connected. The number α is an arbitrary integer within the range that the pointer value Pn can take, and is set to a number that does not adversely affect the sound quality during sound reproduction.

  The ADPCM decoder in FIG. 1B has a memory (MEM) 30 that stores an ADPCM value Dn, and an adapting unit 31 is connected to the output side. In the first embodiment, a comparator (CMP) 32 is newly added to the output side of the adaptation unit 31, and a decoder (DEC) 33 is connected to the output side of the comparator 32 and the memory 30. Yes. The comparator 32 detects the thinned data portion by the comparison operation processing based on the pointer values Pn,... And the quantization width Δn given from the adaptation unit 31, and interpolates the data at the time of decoding. Is given to the decoder 33.

An adder 34 that outputs a reproduction value Yn is connected to the output side of the encoder 33, and a delay unit (Z- ¹ ) 35 is feedback-connected to the adder 34. Further, a D / A converter 36 for converting the reproduction value Yn into an analog audio signal is connected to the output side of the adder 34.

(Encoding method of ADPCM encoder in FIG. 1A)
FIG. 5 is a waveform diagram showing how the ADPCM values in FIG.

  With the configuration of the ADPCM encoder in FIG. 1A, data compression can be performed more efficiently than the conventional ADPCM method. Processing procedures (1) to (7) at this time will be described below.

  (1) The input analog audio signal (audio waveform) is sampled by the A / D converter 20 and converted to a digital signal (current nth original waveform value Xn) and sent to the differentiators 21 and 27. It is done. The difference unit 21 calculates a difference value dn (= Xn− (Yn) from the value Xn of the original waveform and the reproduction value Yn−1 of the previous (n−1) th ADPCM reproduction waveform output from the delay unit 26. -1)) is obtained and sent to the encoder 22.

  (2) The encoder 22 adaptively quantizes the difference value dn based on the quantization width Δn determined by the pointer value Pn−1 given from the adaptation unit 23 to obtain the current ADPCM value Dn and the pointer value Pn. The data is sent to the adaptation unit 23, the decoder 24, and the comparator 29.

  (3) The decoder 24 inversely quantizes the ADPCM value Dn obtained in the procedure (2) to obtain the change amount qn of the equation (2) and sends it to the adder 25.

  (4) The adder 25 adds the amount of change qn obtained in the procedure (3) and the previous (n−1) th reproduction value Yn−1 delayed by the delay unit 26 to obtain the current ADPCM A reproduction value Yn of the reproduction waveform is obtained and sent to the delay unit 26 and the difference unit 27.

  (5) The subtractor 27 subtracts the reproduction value Yn obtained in step (4) from the original waveform value Xn + 1 to obtain the next (n + 1) th difference value dn + 1. To the generator 28.

  (6) The encoder 28 quantizes the difference value dn + 1 by the quantization width Δn + 1 determined by the pointer value Pn obtained in the procedure (2) to obtain the ADPCM value Dn + 1 and the pointer value Pn + 1. Obtained and sent to the comparator 29.

  (7) The comparator 29 compares the ADPCM value Dn and pointer value Pn output from the encoder 22 with the ADPCM value Dn + 1 and pointer value Pn + 1 output from the encoder 28. As shown in FIG. 5, the ADPCM value Dn is stored in the memory 30 according to the following thinning conditions (A1) to (A3) based on the pointer values Pn−1, Pn, and Pn + 1.

(A1) When α>Pn-1>Pn> Pn + 1
The ADPCM value Dn is not stored in the memory 30. That is, in FIG. 5, the current (n-th) sampling data has a smaller change in quantization width Δn than the preceding and following (n−1, n + 1) sampling data. (In other words, since the sampling data that can be regarded as having a constant quantization width Δn is continuous, the effect on the sound quality is small even if approximation by interpolation is performed). It is not thinned out and stored in the memory 30.

(A2) When α>Pn-1> Pn <Pn + 1, Pn + 1> α
The ADPCM value Dn is stored in the memory 30 without being thinned out. The pointer value Pn is set to Pn-1.

(A3) Other times
The ADPCM value Dn is stored in the memory 30 without being thinned out.

  Thereafter, the subsequent n + 1, n + 2,..., ADPCM values Dn + 1, Dn + 2,... Are determined and stored in the memory 30.

(Decoding method of ADPCM decoder in FIG. 1B)
FIG. 6 is a waveform diagram showing how the ADPCM value is reproduced in FIG.
In the configuration of the ADPCM decoder in FIG. 1B, it is possible to accurately decode the ADPCM value. Processing procedures (1) and (2) at this time will be described below.

  (1) The current (n-th) ADPCM value Dn stored in the memory 30 is read and sent to the adaptation unit 31 and the decoder 33. The adaptation unit 31 obtains a quantization width Δn determined by the current (n-th) pointer value Pn from the ADPCM value Dn, and sends it to the comparator 32.

  (2) The comparator 32, the decoder 33, the adder 34, and the delay unit 35 decode the ADPCM value Dn according to the following interpolation conditions (B1) and (B2).

(B1) When α>Pn−1> Pn In order to detect the thinned data portion (the nth sampling data in FIG. 5), the comparator 32 calculates the number α and the pointer values Pn−1 and Pn. When the interpolation condition (B1) is satisfied, it is determined that this is a thinned data portion, and the quantization width Δn / 2 for interpolation is output and sent to the decoder 33. The decoder 33 adaptively dequantizes the ADPCM value Dn according to Equation (2) using the interpolation quantization width Δn / 2 to obtain the change amount qn.

  That is, as shown in FIG. 6, the decoder 33 sets the quantization width Δn by Δn / 2 between the n−1th sampling data and the nth sampling data for the ADPCM value Dn according to Equation (2). Is interpolated, and the interpolated change amount qn is supplied to the adder 34. The adder 34 adds the previous (n−1) th reproduction value Yn−1 delayed by the delay device 35 and the current (nth) variation qn according to the equation (1) An (nth) reproduction value Yn is obtained and supplied to the D / A converter 36 and the delay unit 35. The D / A converter 36 converts the reproduction value Yn into an analog audio signal and outputs it.

(B2) In other cases The comparator 32 compares the number α with the pointer values Pn−1 and Pn, and outputs the quantization width Δn to the decoder 33 when the interpolation condition (B1) is not satisfied. send. The decoder 33 adaptively dequantizes the ADPCM value Dn according to the equation (2) using the quantization width Δn to obtain the change amount qn. The adder 34 calculates the current reproduction value Yn by adding the previous reproduction value Yn-1 delayed by the delay unit 35 and the current change amount qn according to the equation (1), and calculates D / A This is applied to the converter 36 and the delay device 35. The D / A converter 36 converts the reproduction value Yn into an analog audio signal and outputs it.

  Thereafter, the subsequent reproduction values Yn + 1, Yn + 2,... Are obtained in the same manner, and converted to analog audio signals by the D / A converter 36. Go.

(Effect of Example 1)
According to the first embodiment, in the encoding method using the ADPCM encoder, the comparator 29 determines that the thinning-out condition (A1) is satisfied based on α set to a number that does not adversely affect the sound quality at the time of sound reproduction. Sometimes, the ADPCM value Dn is thinned out and not stored in the memory 30, so that the amount of compressed data stored in the memory 30 can be reduced. Further, in the decoding method by the ADPCM decoder, when it is determined by the comparator 32 that the interpolation condition (B1) is satisfied based on α set to a number that does not adversely affect the sound quality at the time of audio reproduction, the comparator 29 Since the thinned value is interpolated by the decoder 33, the original audio signal can be reproduced with high accuracy. Accordingly, it is possible to increase the data compression amount by suppressing the deterioration of the sound quality at the time of reproducing the sound, compared with the conventional case.

(Configuration of Example 2)
FIGS. 7A and 7B are diagrams showing a data compression method in the ADPCM encoding method and decoding method according to Embodiment 2 of the present invention. FIG. 7A is a schematic diagram of the ADPCM encoder. The configuration diagram and FIG. 5B are schematic configuration diagrams of the ADPCM decoder. In FIG. 7, elements common to the elements in FIG. 1 illustrating the first embodiment are denoted by common reference numerals.

  In the ADPCM encoder shown in FIG. 7 (a), instead of the subtracter 27, encoder (ENC) 28, and comparator (CMP) 29 of the ADPCM encoder shown in FIG. (CMP) 29A is provided.

  The comparator 29A is connected between the output side of the subtractor 21 and the encoder (ENC) 22 and the input side of the memory (MEM) 30, and the difference value dn output from the subtractor 21 is encoded. The ADPCM value Dn output from the device 22 is input, the data is thinned out by comparison processing of the difference value dn and a predetermined value (for example, the value β), and the ADPCM value Dn ′ is stored in the memory 30. . The value β is an arbitrary value that can be taken by the difference value dn, and is set to a value that does not adversely affect the sound quality during sound reproduction. Other configurations are the same as those in FIG.

  In the ADPCM decoder of FIG. 7B, a comparator (CMP) 32A having a different configuration is provided instead of the comparator (CMP) 32 of the ADPCM decoder of FIG. 1B.

  The comparator 32A is connected between the output side of the memory (MEM) 30 and the adapting unit 31 and the input side of the decoder (DEC) 33, and the ADPCM value Dn ′ read from the memory 30; The quantization width Δn output from the adapting unit 31 is input, the thinned data portion is detected by the comparison calculation process based on a predetermined value (for example, the value β), and the data at the time of decoding is interpolated Therefore, the quantization width Δn is provided to the decoder 33. The value β is set to a value that does not adversely affect the sound quality during sound reproduction. Other configurations are the same as those in FIG.

(Encoding method of ADPCM encoder in FIG. 7A)
FIG. 8 is a waveform diagram showing how the ADPCM values in FIG.

  In the configuration of the ADPCM encoder in FIG. 7A, data compression can be performed more efficiently than the conventional ADPCM method by using the value Xn and the value β of the original waveform. Processing procedures (1) to (6) at this time will be described below.

  (1) The input analog audio signal (audio waveform) is sampled by the A / D converter 20, converted into a digital signal (current nth original waveform value Xn), and sent to the subtractor 21. The difference unit 21 calculates a difference value dn (= Xn− (Yn) from the value Xn of the original waveform and the reproduction value Yn−1 of the previous (n−1) th ADPCM reproduction waveform output from the delay unit 26. -1)) is obtained and sent to the encoder 22 and the comparator 29A.

  (2) The encoder 22 adaptively quantizes the difference value dn based on the quantization width Δn determined by the pointer value Pn given from the adaptation unit 23 to obtain the current ADPCM value Dn. To the converter 24 and the comparator 29A.

  (3) The comparator 29A inputs the difference value dn obtained in the procedure (1) and the ADPCM value Dn obtained in the procedure (2), performs comparison operation processing, and performs the following (A1), (A2 ) The ADPCM value Dn ′ is obtained according to the thinning condition.

(A1) When dn ≧ β
1-bit data 0 is added to the ADPCM value Dn to obtain the ADPCM value Dn ′ = (Dn, 0).

(A2) When dn <β
1-bit data 1 is added to the ADPCM value Dn to obtain an ADPCM value Dn ′ = (Dn, 1).

  (4) The ADPCM value Dn ′ obtained in the procedure (3) is thinned according to the following thinning conditions (A3) and (A4) and stored in the memory 30 as shown in FIG.

(A3) When Dn ′ = (Dn, 1), Dn-1 ′ = (Dn−1,1), and Dn−1 ′ is stored in the memory 30
The ADPCM value Dn ′ is thinned out and not stored in the memory 30. That is, in FIG. 8, the current (n-th) sampling data has a data change smaller than a predetermined value (value β) as compared to the previous (n−1) -th sampling data (that is, after (n + 1) (Th) data change is small, and even if approximated by interpolation, the influence on the sound quality is small). Therefore, the current (nth) sampling data is thinned out and not stored in the memory 30.

(A4) Other times
The ADPCM value Dn ′ is stored in the memory 30 without being thinned out.

  (5) The decoder 24 inversely quantizes the ADPCM value Dn obtained in the procedure (2) to obtain the change amount qn of the equation (2) and sends it to the adder 25.

  (6) The adder 25 adds the change amount qn obtained in the procedure (5) and the previous (n−1) th reproduction value Yn−1 delayed by the delay unit 26, and adds the current ADPCM. A reproduction value Yn of the reproduction waveform is obtained and sent to the delay unit 26 and the difference unit 27.

  Thereafter, the subsequent n + 1, n + 2,..., ADPCM values Dn + 1, Dn + 2,... Are determined and stored in the memory 30.

(Decoding method of ADPCM decoder in FIG. 7B)
FIG. 9 is a waveform diagram showing how the ADPCM value is reproduced in FIG.
With the configuration of the ADPCM decoder in FIG. 7B, it is possible to accurately decode the ADPCM value. Processing procedures (1) and (2) at this time will be described below.

  (1) The current (n-th) ADPCM value Dn ′ = (Dn, 0 or 1) stored in the memory 30 is read and sent to the adaptation unit 31 and the comparator 32A. The adaptation unit 31 obtains a quantization width Δn determined by the current (n-th) pointer value Pn from the ADPCM value Dn, and sends it to the comparator 32.

  (2) The comparator 32A, the decoder 33, the adder 34, and the delay unit 35 decode the ADPCM value Dn according to the following interpolation conditions (B1) and (B2).

(B1) When Dn '= (Dn, 1) and Dn-1' = (Dn-1,1), the comparator 32A detects the thinned data portion (the nth sampling data in FIG. 8). Therefore, it is determined whether or not the interpolation condition (B1) is satisfied by performing a comparison calculation process. If the interpolation condition (B1) is satisfied, it is determined that this is a thinned data portion, and the interpolation is performed. Is output to the decoder 33. The decoder 33 adaptively dequantizes the ADPCM value Dn according to Equation (2) using the interpolation quantization width Δn / 2 to obtain the change amount qn.

  That is, as shown in FIG. 9, the decoder 33 obtains the value obtained by setting the quantization width Δn as Δn / 2 by the equation (2) between the n−1th sampling and the nth sampling with respect to the ADPCM value Dn. Interpolation is performed, and the interpolated change amount qn is supplied to the adder 34. The adder 34 adds the previous (n−1) th reproduction value Yn−1 delayed by the delay device 35 and the current (nth) variation qn according to the equation (1) An (nth) reproduction value Yn is obtained and supplied to the D / A converter 36 and the delay unit 35. The D / A converter 36 converts the reproduction value Yn into an analog audio signal and outputs it.

(B2) Other times When the comparator 32A determines that the interpolation condition (B1) is not satisfied, the comparator 32A outputs the quantization width Δn and sends it to the decoder 33. The decoder 33 adaptively dequantizes the ADPCM value Dn according to the equation (2) using the quantization width Δn to obtain the change amount qn. The adder 34 calculates the current reproduction value Yn by adding the previous reproduction value Yn-1 delayed by the delay unit 35 and the current change amount qn according to the equation (1), and calculates D / A This is applied to the converter 36 and the delay device 35. The D / A converter 36 converts the reproduction value Yn into an analog audio signal and outputs it.

  Thereafter, the subsequent reproduction values Yn + 1, Yn + 2,... Are obtained in the same manner, and converted to analog audio signals by the D / A converter 36. Go.

(Effect of Example 2)
According to the second embodiment, in the encoding method using the ADPCM encoder, the comparator 29A determines that the thinning-out condition (A3) is satisfied based on β set to a value that does not adversely affect the sound quality at the time of sound reproduction. Sometimes, the ADPCM value Dn ′ is thinned out and not stored in the memory 30, so that the amount of compressed data stored in the memory 30 can be reduced. Further, in the decoding method by the ADPCM decoder, when it is determined by the comparator 32A that the interpolation condition (B1) is satisfied based on β set to a value that does not adversely affect the sound quality at the time of audio reproduction, the comparator 29A Since the thinned value is interpolated by the decoder 33, the original audio signal can be reproduced with high accuracy. Accordingly, in substantially the same manner as in the first embodiment, it is possible to increase the data compression amount by suppressing the deterioration of the sound quality at the time of reproducing the sound as compared with the conventional example.

  The present invention is not limited to the first and second embodiments. For example, the ADPCM encoder shown in FIGS. 1 (a) and 7 (a) and the ADPCM decoder shown in FIGS. 1 (b) and 7 (b) In addition to the illustrated individual circuits, various usage forms and modifications are possible, such as implementation by program control using a processor.

It is a figure which shows the data compression method in the encoding method and decoding method of the ADPCM system which shows Example 1 of this invention. It is a wave form diagram which shows the principle of operation of the conventional ADPCM system. It is a schematic block diagram which shows the conventional ADPCM encoder. It is a schematic block diagram which shows the conventional ADPCM decoder. It is a wave form diagram which shows the mode of the storing to the memory 30 of the ADPCM value in Fig.1 (a). It is a wave form diagram which shows the mode of reproduction | regeneration of the ADPCM value in FIG.1 (b). It is a figure which shows the data compression method in the encoding method and decoding method of the ADPCM system which shows Example 2 of this invention. It is a wave form diagram which shows the mode of the storing to the memory 30 of the ADPCM value in Fig.7 (a). It is a wave form diagram which shows the mode of reproduction | regeneration of the ADPCM value in FIG.7 (b).

Explanation of symbols

20 A / D converter 21, 27 Differentiator 22, 28 Encoder 23, 31 Adaptation unit 24, 33 Decoder 25, 34 Adder 26, 35 Delay device 29, 29A, 32, 32A Comparator 30 Memory 36 D / A converter

Claims (9)

Find the difference value dn between the original waveform value Xn of the digitized audio signal and the playback value Yn-1 of the previous playback waveform,
Quantize the difference value dn with a quantization width Δn determined by the previous pointer value Pn-1 to obtain an adaptive differential pulse code modulation value Dn and a pointer value Pn,
Find the previous pointer value Pn-1 and the next pointer value Pn + 1 of the pointer value Pn,
Based on the pointer values Pn-1, Pn, Pn + 1, when the change in the quantization width Δn is smaller than a predetermined value compared to the preceding and succeeding data, the adaptive differential pulse code modulation value Dn is thinned out from the predetermined value. The coding method of the adaptive differential pulse code modulation method, wherein the adaptive differential pulse code modulation value Dn is stored in a memory as it is when it is large.   The adaptive differential pulse code modulation value Dn is thinned out when α> Pn-1> Pn> Pn + 1, where α is the predetermined value in a range that the pointer value Pn can take. The encoding method of the adaptive differential pulse code modulation system according to Item 1. The quantization width Δn determined by the pointer value Pn is obtained from the adaptive differential pulse code modulation value Dn stored in the memory according to claim 1 or 2,
Find the pointer value Pn-1 immediately before the pointer value Pn,
When the pointer values Pn-1, Pn are smaller than a predetermined value, the adaptive differential pulse code modulation value Dn is dequantized and interpolated by a quantization width smaller than the quantization width Δn, and the adaptive difference is characterized by A decoding method using a pulse code modulation method.   When the predetermined value in the range that the pointer value Pn can take is α, and α> Pn-1> Pn, the adaptive differential pulse code modulation value Dn is set by Δn / 2 which is the small quantization width. 4. The decoding method according to claim 3, wherein the interpolation is performed by inverse quantization. Find the difference value dn between the original waveform value Xn of the digitized audio signal and the playback value Yn-1 of the previous playback waveform,
Quantize the difference value dn with a quantization width Δn determined by the pointer value Pn to obtain an adaptive differential pulse code modulation value Dn,
The adaptive differential pulse code modulation value Dn is thinned out when the difference value dn is smaller than a predetermined value, and the adaptive differential pulse code modulation value Dn is stored in a memory as it is when larger than the predetermined value. A differential pulse code modulation encoding method.   6. The adaptive differential pulse code modulation method according to claim 5, wherein when the predetermined value that can be taken by the differential value dn is β, the adaptive differential pulse code modulation value Dn is thinned out when dn <β. Encoding method.  7. The coding method of the adaptive differential pulse code modulation method according to claim 6, wherein when dn <β, 1 bit is added to the adaptive differential pulse code modulation value Dn and data is thinned out. The quantization width Δn is obtained from the adaptive differential pulse code modulation value Dn stored in the memory according to claim 5,
When the adaptive differential pulse code modulation value Dn is thinned,
A decoding method of an adaptive differential pulse code modulation method, wherein the adaptive differential pulse code modulation value Dn is dequantized and interpolated with a quantization width smaller than the quantization width Δn.   9. The decoding method of the adaptive differential pulse code modulation system according to claim 8, wherein the adaptive differential pulse code modulation value Dn is dequantized and interpolated by Δn / 2 which is the small quantization width.

Images