JP3406275B2 - Digital signal encoding method, digital signal decoding method, these devices and their respective program recording media - Google Patents

Abstract

PROBLEM TO BE SOLVED: To compress a digital input signal at a high compression rate and to completely reproduce the signal. SOLUTION: This reversible coding method includes a process, where a digital input signal is converted into a bit stream in a code absolute value expression for each frame, a process where the bit stream is separated into each bit (110), a process where the bits are connected by each bit position in time sequence (160), and a process where reversible coding is applied to each bit stream obtained by the connection (150). The reversible decoding method includes a process, where reversible decoding is applied to each frame of reversible coding series (210), a process where each bit obtained by the reversible decoding is separated into respective bits, a process where each bit is connected by each same time in the order of bit location (220), and a process where frames obtained by the connection are connected sequentially (250).

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an encoding method for eliminating the number of bits representing various digital signals (for example, audio signals such as voice and music and image signals), a decoding method for reproducing the code, The present invention relates to these devices and program recording media.

[0002]

2. Description of the Related Art As a method of compressing information such as voice and image, there are lossy compression coding which allows distortion and lossless compression coding which does not allow distortion. For lossy compression, ITU
Various systems are known as standards for -T and ISO / IEC MPEG. By using these methods, it is possible to compress the original digital signal to 1/10 or less while suppressing the distortion to a slight extent. However, the distortion depends on the coding condition and the input signal, and depending on the application, the deterioration of the reproduced signal may be a problem. On the other hand, as a lossless compression method capable of completely reproducing the original signal, universal compression encoding which is often used for compression of computer files and texts is known. This is compression coding with reference to the statistics of the input sequence. It can compress any signal, and can compress text etc. to about 1/2, but directly to audio or image data. Even if it is applied, the compression effect remains only about 20%.

[0003]

As described above, in the past, lossy coding and lossless coding were selectively used depending on the type and use of information to be compressed. In this case, 1
Since it is necessary to prepare two types of compressed files for one piece of information, the compression efficiency as a whole does not increase and the handling is inconvenient. An object of the present invention is to provide an encoding method for compressing a digital signal with high compression efficiency without distortion, and a decoding method for restoring a signal compressed by the method to an original signal.

[0004]

SUMMARY OF THE INVENTION In the present invention, distortion-tolerant coding is used to quantize a digital input signal with a small number of bits. Then, an error signal between the quantized signal and the original digital input signal (a signal having a reduced amplitude)
Are compressed using distortion-free coding. As a result, the point that the compression rate can be increased without the distortion finally occurring is a characteristic which is not present in the conventional encoding that uses the lossless encoding and the lossy encoding independently.

[0005]

BEST MODE FOR CARRYING OUT THE INVENTION The following embodiments do not limit the claimed invention. Further, not all combinations of the features described in the embodiments are required to achieve the object. An embodiment of the present invention will be described below with reference to the drawings.

§1. First Embodiment FIG. 1 is a block diagram showing a configuration example of an encoding device and a decoding device according to the first embodiment of the present invention. In the encoding device shown in FIG. 1, a time series of digital input signals (hereinafter referred to as “digital input signal series”) is input from an input terminal 100. And the frame division unit 1
10 designates the digital input signal sequence as, for example, 102
It is sequentially divided into frame units made up of four digital input signals (ie, samples of 1024 points).

Next, in order to improve the efficiency of lossless compression coding by the lossless coding unit 150, the reordering unit 160
Each digital input signal (ie bit string) in the frame
The bit of is replaced. Details of the processing performed by the reordering unit 160 will be described below.

FIG. 2 is an explanatory diagram showing an example of the processing of the reordering section 160 in the first embodiment. 2A
Indicates an input to the reordering unit 160 (that is, a digital input signal for one frame). In each digital input signal, the amplitude (here, a positive or negative integer) of the digital input signal is represented by a two's complement representation.

First, the reordering unit 160 converts each digital input signal in a frame from a bit string represented by 2's complement to a bit string represented by a code absolute value (see B in FIG. 2). In the converted digital input signal, MS
B (Most Significant Bit) to 2nd LS
B indicates the absolute value of the amplitude, and LSB (Least Significant
Bit: The least significant bit) indicates the sign of the amplitude.

Next, the reordering unit 160 decomposes the converted digital input signal into individual bits. Finally,
The reordering unit 160 assigns each bit generated by the decomposition to the bit position (ie, MSB, second MS) of each bit.
(B, ..., LSB), that is, in the horizontal direction of FIG. 2 (see C of FIG. 2). The reordering section 160 is
Not all of the bits generated by decomposition but some of the bits generated by decomposition may be combined. Further, the reordering unit 160 may combine the bits generated by the decomposition not at each bit position but at each bit position.

FIG. 3 is an explanatory diagram showing an example of the output of the rearrangement section 160 in the first embodiment (hereinafter referred to as "converted digital input signal"). In FIG. 3, each bit string in the horizontal direction (that is, a bit string composed of 1024 bits at the same bit position) is referred to as a “horizontal bit string”.

Next, the lossless coding unit 150 performs lossless compression coding on the output of the reordering unit 160. Lossless encoding unit 1
Details of the processing by 50 will be described below.

First, the lossless encoding unit 150 encodes horizontal bit strings from the MSB to the jMSB (for example, the 4th MSB) by the following methods. If all the 1024 bits forming the horizontal bit string are "0", the horizontal bit string is encoded into "0". If one bit is “1” among the 1024 bits forming the horizontal bit string, the horizontal bit string is
A bit string indicating the position of the bit that is "1" (10 bits as an example) is encoded into a bit string that is concatenated after "10". If two bits are “1” among the 1024 bits forming the horizontal bit string, the horizontal bit string is
A bit string indicating the position of the bit that is "1" (10 bits x 2 as an example) is encoded into a bit string that is concatenated after "110". If three or more of the 1024 bits forming the horizontal bit string are "1", the horizontal bit string is connected after "111".

Further, the lossless encoding unit 150 has the (j +) th
1) Each horizontal bit string from MSB to LSB is encoded. In this case, the lossless encoding unit 150 may encode the horizontal bit string by any one of the above methods,
Alternatively, each digital input signal may be recombined (that is, in the vertical direction of FIG. 3) and then encoded by the unit of the digital input signal using a generally well-known encoding method.

The lossless encoding unit 150 may perform lossless compression encoding using a case where there is a continuous sequence or a sequence which appears frequently, in addition to the above-mentioned encoding method. Entropy coding such as Huffman coding and arithmetic coding is an example. The compression efficiency is also improved by applying universal encoding for reversibly compressing text or the like to the output of the reordering unit 160.

By the above processing, the coding apparatus shown in FIG.
Output (e).

On the other hand, in the decoding device shown in FIG. 1, first, the lossless decoding unit 210 first includes the lossless encoding unit 15.
The lossless compression code I (e) is decoded by performing a process (that is, decoding) that is the reverse of 0. Next, the reordering unit 220 performs a process reverse to that of the reordering unit 160 to sequentially output the digital input signal in frame units. Finally, the frame synthesizing unit 250 reproduces the original digital input signal sequence by sequentially connecting the outputs of the reordering unit 220. Through the above processing, the digital input signal sequence is output from the output terminal 260.

§2. Second Embodiment FIG. 4 is a block diagram showing a configuration example of an encoding device and a decoding device according to the second embodiment of the present invention. 4, parts corresponding to the respective parts in FIG. 1 are designated by the same reference numerals, and the description thereof will be omitted. In the encoding device shown in FIG. 4, first, the frame division unit 110 divides the digital input signal sequence from the input terminal 100 into frame units made up of, for example, 1024 digital input signals (ie, 1024 sample points). Divide sequentially. Next, the lossy quantization unit 120 compression-codes the output of the frame division unit 110. This encoding may be any method suitable for the input signal as long as the original digital input signal can be reproduced to some extent at the time of decoding. For example, if the digital input signal is voice, ITU-T voice coding or the like can be used, if it is music, MPEG or Twin VQ or the like can be used, and if it is video, MPEG or the like can be used.
The output of the lossy quantization unit 120 is referred to as “lossy compression code I (n)”.

Next, the inverse quantizer 130 having the same structure as the decoder corresponding to the irreversible quantizer 120 (that is, the inverse quantizer 230) outputs the locally reproduced signal from the irreversible compression code I (n). To generate. Then, the subtracting section 140 obtains an error signal between the local reproduction signal and the original digital input signal. Normal,
The amplitude of this error signal is significantly smaller than the amplitude of the original digital input signal. Therefore, lossless compression coding of the error signal is easier than lossless compression coding of the digital input signal. If the error signal can be reproduced without distortion, the original digital input signal can be reproduced without distortion by combining the lossless compression-encoded error signal and the lossy compression-encoded original digital input signal. Therefore, by utilizing this fact (the error signal has a small amplitude and is easily reversibly compressed and encoded), the reversible encoding unit 150 reversibly compression-encodes the error signal.

At this time, in order to improve the efficiency of the lossless compression coding by the lossless coding unit 150, the reordering unit 160
Replaces the bits of the error signal (ie, bit string). Details of the processing performed by the reordering unit 160 will be described below. FIG. 5 shows the reordering portion 16 according to the second embodiment.
It is explanatory drawing which shows an example of the process of 0. In the digital input signal (see A in FIG. 5), positive and negative integers are expressed in 2's complement notation. FIG. 5B shows an error signal between the digital input signal shown in FIG. 5A and the local reproduction signal corresponding to the digital input signal. The reordering unit 160 converts the error signal (that is, the bit string) from the bit string represented by 2's complement into the bit string represented by the sign absolute value (see C in FIG. 5). Next, the reordering unit 160 decomposes the converted error signal into individual bits. Finally, the reordering unit 160 replaces each bit generated by the decomposition with
Bit position of each bit (ie MSB, second MSB, ...
, LSB), that is, in the horizontal direction of FIG. 5 (see D of FIG. 5). FIG. 6 is an explanatory diagram showing an example of the output (hereinafter, referred to as “conversion error signal”) of the reordering unit 160 in the second embodiment. In FIG. 6, each bit string in the horizontal direction (that is, a bit string composed of 1024 bits at the same bit position) is referred to as a “horizontal bit string”. With the above arrangement, the value of each error signal remains unchanged. However, since the error signal has a small amplitude, in the output of the reordering unit 160, all the upper bits are often “0” as shown in a concrete example in FIG. As a result, the continuous loss of “0” improves the lossless compression coding efficiency of the error signal. It can be said that the above replacement is to bit-slice the error signal (that is, the vertical bit string) in the horizontal direction.

Next, the lossless coding unit 150 performs lossless compression coding on the output of the reordering unit 160. Lossless encoding unit 1
Details of the processing by 50 will be described below.

First, the lossless encoding unit 150 encodes horizontal bit strings from the MSB to the jth MSB (for example, the 4th MSB) by the following methods (1) to (3). If all the 1024 bits forming the horizontal bit string are "0", the horizontal bit string is encoded into "0". If one bit is “1” among the 1024 bits forming the horizontal bit string, the horizontal bit string is
A bit string indicating the position of the bit that is "1" (10 bits as an example) is encoded into a bit string that is concatenated after "10". If two bits are “1” among the 1024 bits forming the horizontal bit string, the horizontal bit string is
A bit string indicating the position of the bit that is "1" (10 bits x 2 as an example) is encoded into a bit string that is concatenated after "110". If three or more of the 1024 bits forming the horizontal bit string are "1", the horizontal bit string is connected after "111".

Further, the lossless encoding unit 150 has the (j +) th
1) Each horizontal bit string from MSB to LSB is encoded. In this case, the lossless encoding unit 150 may encode the horizontal bit string by any one of the above methods,
Alternatively, each error signal may be re-combined (that is, in the vertical direction in FIG. 6) and then encoded by the error signal unit using a generally well-known encoding method.

[According to the above lossless compression encoding, the MSB
Even if there are many digital input signals having "1", the amplitude of the error signal corresponding to the digital input signals is 20 [d
B] It can be expected from FIG. 6 that "when it becomes smaller, there is almost no" 1 "in the MSB to the third MSB of the error signal." In this case, by performing the reordering described above, it is possible to reduce approximately 3/16 of the number of bits of the error signal as compared with the case where the reordering is not performed.

The lossless encoding unit 150 may perform lossless compression encoding using a case where there is a continuous sequence or a sequence which appears frequently, in addition to the above-mentioned encoding method. Entropy coding such as Huffman coding and arithmetic coding is an example. The compression efficiency is also improved by applying universal encoding for reversibly compressing text or the like to the output of the reordering unit 160.

By the above processing, the coding apparatus shown in FIG. 4 has the lossy compression code I (n) from the lossy quantization unit 120 and the lossless compression code I from the lossless coding unit 150.
(E) and are output.

On the other hand, in the decoding device shown in FIG. 4, the lossless decoding unit 210 decodes the lossless compression code I (e). Then, the reordering unit 220 performs a process reverse to that of the reordering unit 160 to sequentially output the error signal in frame units. Also, the inverse quantization unit 230
Decodes the lossy compression code I (n). Next, the addition unit 240 replaces the output of the inverse quantization unit 230 with the replacement unit 2.
And the output of 30 are added. Finally, the frame synthesis unit 25
0 reproduces the original digital input signal sequence by sequentially concatenating the outputs of the adder 240. Through the above processing, the digital input signal sequence is output to the output terminal 2
It is output from 60.

The lossy quantizer 120 in the coding apparatus shown in FIG. 4 can be constructed by, for example, a transform coding method. This transform coding method is disclosed, for example, in Japanese Patent Laid-Open No. Hei 8
-44399. This example will be briefly described with reference to FIG. From the output of the frame dividing unit 110, the frame dividing unit 14 outputs the past 2N for every N samples.
Extract the sample. This sequence consisting of 2N samples is a Lapted Orthogonal Transform (LOT).
Transform) Processing frame. Time window 15
However, a time window is applied to this LOT processing frame. On the other hand, the output of the time windowing unit 15 is an Nth-order MDCT (Modified Discrete Cosine) which is a type of superposition orthogonal transform.
Transform: Modified Discrete Cosine Transform) section 16
A modified discrete cosine transform is applied to the frequency domain coefficients. On the other hand, the output of the time windowing unit 15 is subjected to linear prediction analysis by the linear prediction analysis unit 17. As a result, the P-th order prediction coefficient α
0, ..., αP are obtained. The quantizing unit 18 converts the prediction coefficients α 0, ..., αP into, for example, an LSP parameter or a k parameter and then quantizes the predictive coefficient α 0, ...
To get Here, the "spectrum envelop (spectrum envel
ope) ”means the amplitude envelope of the MDCT coefficients.

The spectrum outline calculation unit 21 obtains the spectrum outline of the prediction coefficients α 0, ..., αP. The normalization unit 22 divides (ie, normalizes) the spectrum amplitude from the MDCT unit 16 for each corresponding sample in the above-mentioned spectrum outline, so that the current frame F is obtained.
The residual coefficient R (F) of Where "residual coefficient (resi
"dual coefficient)" means the MDCT coefficient flattened in the spectral shape. The weight calculation unit 24 multiplies the spectrum outline from the spectrum outline calculation unit 21 and the residual coefficient outline ER (F) from the residual outline calculation unit 23 for each corresponding sample to obtain the weighting coefficient. , WN (represented by the vector W) are obtained.
Here, the “envelope of residual coeff
icient) "means a rough analysis of the residual coefficient on the frequency axis. Then, the weight calculator 24
Supplies the weighting coefficient W to the quantizer 25.

The residual outline normalization unit 26 divides the residual coefficient R (F) of the current frame F from the normalization unit 22 by the residual coefficient outline from the residual coefficient outline calculation unit 23 ( That is, by normalizing), the fine structure coefficient is obtained. here,
“Fine structure coefficient” is
The MDCT coefficient obtained by further flattening the flattened MDCT coefficient by the residual coefficient outline. Power normalization unit 2
7 is a normalized gain g which is the square root of the average value of the amplitude or the average value of the power of the fine structure coefficient of the current frame.
It is divided (i.e., normalized) by (F), and the result is supplied to the quantization unit 25 as a normalized fine structure coefficient X (F) = (x1, ..., xN). The power normalization unit 27 gives the normalization gain g (F) to the denormalization unit 31. Also,
The power normalization unit 27 quantizes the normalized gain g (F), and outputs the result as an index IG.

The quantizing unit 25 calculates the normalized fine structure coefficient X.
After weighting (F) with the weighting coefficient W, vector quantization is performed to obtain a quantized small sequence C.
(M) is obtained. Here, the quantized subsequence C (m) is M vectors C (m1), C (m2), ..., C (mM).
) Is a series of elements. The denormalization unit 31 denormalizes the quantized subsequence C (m) by the normalization gain g (F) from the power normalization unit 27, and further, the residual coefficient from the residual outline calculation unit 23. The quantized residual coefficient Rq (F) is reproduced by multiplying by the rough shape. The residual outline calculation unit 23 obtains an outline of the quantized residual coefficient Rq (F).

The index IP indicating the quantized value of the linear prediction coefficient, the index IG indicating the quantized value of the power normalization gain of the fine structure coefficient, and the index Im indicating the quantized value of the fine structure coefficient are irreversible. Compression code I
(N) is output from the lossy quantization unit 120.

In the inverse quantization unit 130, the normalized fine structure coefficient corresponding to the index Im and the normalized gain corresponding to the index IG are input to the power inverse normalization unit 53, respectively. Power denormalization unit 53
Obtains the fine structure coefficient by denormalizing the normalized fine structure coefficient by the normalized gain. The residual outline denormalization unit 54 uses the residual outline E from the residual outline calculation unit 55.
Multiply the fine structure coefficient by R 1 (ie, denormalize)
To reproduce the residual coefficient Rq (F). The residual outline calculating unit 55 calculates the outline of the reproduced residual coefficient Rq (F) by the same method as the residual outline calculating unit 23 of the irreversible quantization unit 120.

On the other hand, the inverse normalization unit 57 multiplies the spectrum outline from the spectrum outline calculation unit 21 by the residual coefficient Rq (F) from the residual outline inverse normalization unit 54 (that is,
Inverse normalization) to reproduce the frequency domain coefficients. The inverse MDCT unit 58 transforms the frequency domain coefficient into a 2N-sample time domain signal (hereinafter, referred to as "inverse LOT processing frame") by performing Nth-order inverse modified discrete cosine transform for each frame. The windowing unit 59 multiplies the time domain signal by a time window for each frame. The frame superposition unit 61 adds the first half N samples of the output of the windowing unit 59 (the current frame for inverse LOT processing of 2N samples) and the second half N samples of the immediately preceding frame to each other, and obtains N as a result. The sample is output to the subtraction unit 140 as the dequantized signal of the current frame.

As an example of the irreversible quantizer 120,
A configuration for dividing in the frequency domain and hierarchically encoding will be briefly described with reference to FIG. This configuration is based on the Tokuhei Hei 8-263
No. 096 publication. The original sound input signal 21 from the frame division unit 110 has the highest frequency of f4,
The sampling frequency is 2f4. The first sample rate converter 221 which is the first band selection means receives the highest frequency f1 and the sampling frequency 2f1 from the original sound input signal 21.
(However, the first band signal 231 of f1 <f2 <f3 <f4) is taken out. The first encoder 241 encodes the first band signal 231 into the first code C1. The first decoder 251 sends the first code C1 to the highest frequency f1
And it is decoded into the decoded signal 121 1 having the sampling frequency 2f 1. The first sample rate converter 261 outputs the decoded signal 121 to the highest frequency f2 and the sampling frequency 2f2.
The first converted decoded signal of On the other hand, the sample rate converter 222, which is the second band selecting means, receives the highest frequency f2 and the sampling frequency 2f from the original sound input signal 21.
2 2nd band signal 232. Second difference circuit 2
82 2 subtracts the first transformed decoded signal from the second band signal 232 2 to obtain the second difference signal 292 1.
The second encoder 242 encodes the second difference signal 292 into a second code C2.

The same processing is performed thereafter. Here, the process of obtaining the third code C3 is performed by i = 3 (i = 2, 3, ..., N,
In this example, up to 4) will be described as an example. I-th (=
2nd) Decoder 25i-1 (252)
The (= second) code Ci-1 (= C2) is converted into the highest frequency fi-1 (= f2) and the sampling frequency 2fi-1.
(= 2f2) i-th (= second) decoded signal. The adder 60i−1 (= 602) is the i−1th
By adding the (= second) decoded signal and the (i-2) th (= first) converted decoded signal, an (i-1) th (second) sum signal is obtained. I-1th (= second) sample rate converter 26i
−1 (= 262) is the i−1th (second) sum signal,
Highest frequency fi (= f3) and sampling frequency 2fi
(= 2f3) i-th (= second) converted decoded signal. On the other hand, the sample rate converter 22i (= 223), which is the i-th (= third) band selecting means, receives from the original sound input signal 21 the i-th signal having the highest frequency fi (= f3) and the sampling frequency 2fi (= 2f3). The (= third) band signal 23i (= 233) is taken out. The i-th (= third) difference circuit 28i (283) shifts from the i-th (= third) band signal 23i (= 233) to the i-1th (= second) signal.
By subtracting the converted decoded signal, the i-th (= third)
The difference signal 293 is obtained. I-th (= third) encoder 24i
(= 243) encodes the i-th (= third) difference signal 293 into the i-th (= third) code Ci (= C3). Note that the i−1th (= second) decoder 25i−1 (= 252)
) And the adder 60i-1 (= 602) and the i-th
(= Second) sample rate converter 26i-1 (= 26
2) is the i-1th (= second) decoding means 40i-1.
(= 402). However, the first decoding means 4
In 01, since the i−2 layer does not exist, the adder 60
0 is omitted. Also, since the band signal of the highest layer, i.e., the i + 1th (= fourth) band signal 234 in this example is the signal of the highest frequency f4, the sample rate converter 224, which is the i + 1th band selecting means, is omitted.

The irreversible quantizer 120 shown in FIG.
It can be applied when the input signal band is divided into n sections and encoded. First to nth (= fourth) codes C1 to Cn (=
C4) is multiplexed for each frame in the multiplexing circuit and then output as the lossy code P (n). In this case, the multiplexing circuit is configured to select and output any of the first to i-th codes. I-th encoder 24i
(I = 1 to n: where n = 4 in FIG. 8) is set so that the compression rate decreases as i increases,
Wideband and high quality coding is possible. If this is satisfied, the encoding method of the encoder may be, for example, all transform encoding methods.

As the encoding method of the first to n-th encoders 241 to 244, the CELP encoding method or the transform encoding method can be used, and the first to n-th encoders 241 to 244 use the same encoding method. May be used, or different encoding methods may be used. The lossy quantization unit 1 shown in FIG.
An outline of the inverse quantization unit 130 when 20 is used will be described with reference to FIG. FIG. 9 shows that when n = 4, that is,
This is the case where the code string C is composed of the 1st to 4th codes C1 to C4. The code separating means 56 converts the code string C into the 1st to 4th codes.
Separation into symbols Cl to C4. Then, the first code C1
Is supplied to the first decoder 571, the second code C2 is supplied to the second decoder 572, and the third code C3 is supplied to the third
It is supplied to the decoder 573, and the fourth code C4 is supplied to the fourth decoder 574. The first decoder 571 decodes the first code C1 into a first decoded signal 581. First
The decoded signal 581 is input to the first sample rate converter 591 as the first decoded output 631. The first sample rate converter 591 has a first decoded output 631.
Is the first converted decoded signal 611 of sampling frequency 2f2.
Convert to. On the other hand, the second decoder 572 decodes the second code C2 into the second decoded signal 582. The second adder 622 2 adds the first transformed decoded signal 611 and the second decoded signal 582 to obtain the second decoded output 632. The second sample rate converter 592
The second decoded output 632 is transformed into a second transformed decoded signal 612 with a sampling frequency 2f3. In the following, the case of i = 3 will be generalized and explained. The i-th (= third) decoder 57i (= 573) is the i-th (= third) code Ci.
(= C3) to the i-th (= third) decoded signal 58i (=
583). I-th (= third) adder 62i
(= 623) is the i−1th (= second) converted decoded signal 6
1i-1 (= 612) and i-th (= third) decoded signal 5
8i (= 583) and the i-th
The (= third) decoded output 63i (= 633) is obtained.
I-th (= third) sample rate converter 59i (= 59)
3) is the i-th (= third) decoded output 63i (= 63)
3) is converted into the i-th (= third) converted decoded signal 61i (= 613) of the sampling frequency 2fi + 1 (= 2f4). I + 1th (= fourth) decoder 57i + 1 (= 5)
74) is the i + 1th (= fourth) code Ci + 1 (= C
4) to the i + 1th (= fourth) decoded signal 58i + 1
(= 584). The i + 1th (= fourth) adder 62i + 1 (= 624) outputs the ith (= third) converted decoded signal 61i (= 613) and the i + 1th (= fourth)
By adding the decoded signal 58i + 1 (= 584), the i + 1th (= fourth) decoded output 63i + 1
(= 634) is obtained. The i + 1th (= fourth) decoded output 63i + 1 (= 634) is supplied to the subtraction unit 140 as a quantized signal.

Next, the experimental results of the second embodiment are shown in FIGS. 3 sampled at 44.1 kHz
Experiments were performed on three different stereo audio signals and three different monaural audio signals sampled at 16 kHz. As the irreversible quantization unit 120,
Twin VQ in ISO / IEC MPEG-4 scalable profile was used. Irreversible quantizer 12
The compression unit bit rate of 0 is 16 kbit / s / ch for 44.1 kHz and 8 kbit / s for 16 kHz. By repeatedly using this quantizer, the sum of the data rate of the MPEG compressed signal and the data rate of the error signal is reduced. As a result, the quantization bit rate is 44.1 kHz.
To 1/44 of the original signal and to 1/32 for 16 kHz. The percentage (%) of the total file size is the file size of the error signal and the compression MP
It is the sum of the size of the EG bitstream. The total file size shown in FIGS. 10 and 11 indicates that the original signal size can be reduced to about 70-50%. It can be seen that, when the lossy quantizer 120 compresses the original signal file size to about 1/10, the total file size is minimized. Also, 44.1k
Hz stereo has a slightly better compression ratio than 16 kHz monaural.

§3. Third Embodiment FIG. 12 is a block diagram showing a configuration example of an encoding device and a decoding device according to the third embodiment of the present invention. Figure 1
In FIG. 2, parts corresponding to the parts in FIGS. 1 and 4 are designated by the same reference numerals, and description thereof will be omitted. The device shown in FIG. 12 has either a configuration equivalent to the device according to the first embodiment (see FIG. 1) or a configuration equivalent to the device according to the second embodiment (see FIG. 4) due to the selection unit 270 and the switching unit 280. Can be switched to

That is, in the encoding device, the selection unit 270
Is input from the frame dividing unit 110 in frame units, and the error signal corresponding to the digital input signal from the subtracting unit 140 in frame units. Then, the selecting section 270 selects and outputs either the digital input signal or the error signal in frame units. The selecting unit 270 also generates, for each output (that is, for each frame), selection information indicating whether the output is a digital input signal or an error signal, and adds the selection information to the output. The selection information is the selection unit 2
Instead of being added to the output of 70, it may be separately sent to the switching unit 280 in synchronization with the output of the selecting unit 270. Hereinafter, the output of the selection unit 270 is subjected to the processing described in the first and second embodiments (that is, reordering → reversible coding → reversible decoding → reverse reordering). As a result, the same signal as the output of the selection unit 270 is output from the reordering unit 220 in frame units.

On the other hand, in the decoding device, the switching unit 280
Extracts selection information from the output of the reordering unit 220. The selection information may be received from the selection unit 270 in synchronization with the output of the reordering unit 220. If the selection information indicates that the output of the rearrangement unit 220 is a digital input signal, the switching unit 2
The 80 sends the output as it is to the frame synthesis section 250. Hereinafter, the output of the switching unit 280 is subjected to the processing (frame synthesis → reproduction signal output) described in the first embodiment. On the other hand, if the selection information indicates that “the output of the reordering unit 220 is an error signal”, the switching unit 280
Sends the output to the frame adder 240. Hereinafter, the output of the switching unit 280 is subjected to the processing described in the second embodiment (addition with the output of the inverse quantization unit 230 → frame synthesis → reproduction signal output).

§4. Fourth Embodiment Before describing the fourth embodiment, problems to be solved by the embodiment will be described. In the lossy compression part (that is, the lossy quantization unit 120, the dequantization unit 130, and the dequantization unit 230) of the second and third embodiments, the known compression format and the decoder are used as they are. There are cases. However, in this case, if the encoding device and the decoding device are implemented on different platforms, the local reproduction signal (that is, the dequantization unit 130) used for generating the error signal is generated.
Output) and the actual dequantized signal (that is, the dequantization unit 23
0 output) does not exactly match. For example, taking the MPEG audio standard as an example, when the digital input signal is 16 bits, the difference between the reference standard output signal for the interconnection test between the encoder and the decoder and the actual output signal of the decoder is Although it is guaranteed that the difference is within "± 1" (that is, an error to the extent that the LSBs do not match), this is not always the case. In this way, when the output of the inverse quantizer 130 and the output of the inverse quantizer 230 do not match, the error signal generated based on the output of the inverse quantizer 130 is added to the inverse quantizer (in the adder 240). The addition to the output of 230 does not guarantee that the original digital input signal will be reproduced without distortion.

Therefore, in the fourth embodiment, the inverse quantizer 1
The absolute value part of the amplitude of the digital input signal is arithmetically shifted to the upper side by the number of bits (probably t bits) based on the difference between the output of 30 and the output of the inverse quantizer 230, and the arithmetically shifted digital input. The signal is irreversibly quantized. Then, after the irreversibly quantized digital input signal is dequantized (in the dequantization unit 130 or the dequantization unit 230), the absolute value part of the amplitude of the dequantized signal is t in the lower direction. Bit arithmetic shift (ie divide by 2t). As a result, the output of the inverse quantization unit 130 and the output of the inverse quantization unit 230 match.

In the case of MPEG audio encoding,
The output of the inverse quantizer 130 and the output of the inverse quantizer 230 are respectively ± 1 compared with the original digital input signal.
Including the error of. As a result, the difference between the output of the inverse quantization unit 130 and the output of the inverse quantization unit 230 is the maximum value “± 2” when one is “+1” and the other is “−1”. . Therefore, in the present embodiment, in order to remove the maximum value of this difference from the digital input signal, t = 3 bits is set, and the digital input signal is arithmetically shifted by 3 bits.
As a matter of course, the value of t is not limited to t = 3, and is determined based on the difference between the output of the inverse quantizer 130 and the output of the inverse quantizer 230 as described above. It

FIG. 13 is a block diagram showing a configuration example of an encoding device and a decoding device according to the fourth embodiment of the present invention. In FIG. 13, parts corresponding to those in FIGS. 1, 4 and 12 are assigned the same reference numerals and explanations thereof are omitted.

In FIG. 13, the frame division unit 110
The digital input signal to the frame, the reproduction output signal from the frame synthesizing unit 250, and the signals processed by the irreversible quantizing unit 120, the dequantizing unit 130, and the dequantizing unit 230 are represented by two's complement representation. , Other signals are represented by sign absolute value representation. Therefore, it is assumed that the conversion from the two's complement representation to the sign absolute value representation and the conversion from the sign absolute value representation to the two's complement representation are appropriately performed (although not shown in FIG. 13). .

In the encoding device shown in FIG. 13, first,
The frame division unit 110 converts the digital input signal sequence into
The frame is sequentially divided. Bit shift unit 290
, The absolute value part of the amplitude of each digital input signal is arithmetically shifted in the upper direction by t bits, and the lower t bits of the absolute value part are set to "0". The lossy quantizer 120
The output of the bit shift unit 290 is lossy compression-encoded. The inverse quantization unit 130 inversely quantizes the output of the irreversible quantization unit 120 to generate a local reproduction signal. The bit shift unit 300 rounds off the lower t bits of the absolute value part of the amplitude of the local reproduction signal, and then arithmetically shifts the lower t bits by t bits. The subtraction unit 140 obtains an error signal between the output of the frame division unit 110 and the output of the bit shift unit 300. After that, the reordering unit 160 and the lossless encoding unit 150 perform the same processing on the error signal as the reordering unit 160 and the lossless encoding unit 150 of the second embodiment. As a result, the lossless encoding unit 150
Outputs a lossless compression code of the error signal.

On the other hand, the lossless encoding unit 310 losslessly compresses and encodes the upper t bits of the absolute value part of the amplitude of each digital input signal. Since each digital input signal is represented by a sign absolute value expression, the upper t bits are often "0". Therefore, the lossless encoding unit 310 can improve the compression efficiency by performing the same encoding process as the lossless encoding unit 150.

By the above processing, the coding apparatus shown in FIG. 13 has the lossy compression code from the lossy quantization unit 120, the lossless compression code (of the error signal) from the lossless coding unit 150, and the lossless The lossless compression code (higher t bits of the digital input signal) from the encoding unit 310 is output.

On the other hand, in the decoding device shown in FIG. 13, the lossless decoding unit 210 and the reordering unit 220 are included.
However, the same processing as the lossless decoding section 210 and the reordering section 220 of the second embodiment is performed on the lossless compression code (of the error signal) from the lossless encoding section 150. On the other hand, the inverse quantizer 230 inversely quantizes the output of the lossy quantizer 120 to generate a local reproduction signal. The bit shift unit 320 rounds off the lower t bits of the absolute value part of the amplitude of the local reproduction signal, and then arithmetically shifts the lower t bits by t bits. The addition unit 240 adds the output of the reordering unit 220 and the output of the bit shift unit 320.

On the other hand, the lossless decoding section 330 losslessly decodes the output of the lossless encoding section 310 to output the upper t bits of the absolute value part of the amplitude of the original digital input signal in frame units. The addition unit 340 connects the output of the lossless decoding unit 330 to the upper bit side of the output of the addition unit 240. Finally, the frame synthesis unit 250 reproduces the original digital input signal sequence by sequentially concatenating the outputs of the addition unit 340. Through the above processing, the digital input signal sequence is output from the output terminal 260.

§5. Supplement The functions of each unit shown in each of the above-described embodiments can also be realized by causing a computer to read and execute a program.

Although the devices shown in FIGS. 4, 12 and 13 have only one type of lossy quantization means (ie, the lossy quantization unit 120, the dequantization unit 130, the dequantization unit). However, the present invention is not limited to this, and includes a plurality of types of lossy quantization methods and a plurality of lossy quantization conditions. (For example, a method and conditions corresponding to inverse quantization that can be executed by the decoding device)
It may be selected for each frame). In that case,
In the third embodiment, the information indicating the selected method and condition may be added to the selection information and sent to the decoding device.

[0055]

According to the present invention, the original digital signal can be completely restored, and at the same time, a high compression rate can be obtained. Further, since the frame is divided, the signal can be completely reproduced even in the middle of the compression code sequence by using a synchronization word or the like. In addition, it has been conventionally practiced to record both the original sound file of music and the compressed code file in a CD-ROM, provide the compressed code file free of charge, and provide the original sound file only to the person who paid the fee. It is being appreciated. According to the present invention, the compression code and the error signal are recorded, the compression code is provided free of charge, and the original sound can be reproduced by providing the compression code and the error signal to the person who paid the fee. can do. With this configuration, the data amount of the sum of the compression code and the error signal is significantly smaller than the data amount of the sum of the original sound and the compression code, and therefore the data can be recorded on a storage medium having a small storage capacity. Although there is a large amount of processing in irreversible quantization and dequantization, hardware that executes these processings at high speed (for example, MPE
Since a variety of G encoding / decoding processors) and software are commercially available and can be obtained at low cost, a large amount of processing does not matter.

As described above, according to the present invention, Twin VQ, MP
By combining with the sound compression encoding such as 3, MPEG-4, the highly efficient lossy encoding and the lossless lossless encoding can coexist without waste. for that reason,
INDUSTRIAL APPLICABILITY The present invention is particularly effective in accumulating and transmitting musical tone signals, and can be applied to a wide range of applications including music network distribution and portable music listening equipment.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration example of an encoding device and a decoding device according to a first embodiment of the present invention.

FIG. 2 is an explanatory diagram illustrating an example of processing of a reordering unit 160 according to the first embodiment.

FIG. 3 is an explanatory diagram illustrating an example of an output of the reordering unit 160 according to the first embodiment.

FIG. 4 is a block diagram showing a configuration example of an encoding device and a decoding device according to a second embodiment of the present invention.

FIG. 5 is an explanatory diagram illustrating an example of processing of a reordering unit 160 according to the second embodiment.

FIG. 6 is an explanatory diagram showing an example of the output of the reordering unit 160 in the second embodiment.

FIG. 7 is a lossy quantization unit 120 using a transform coding method.
FIG. 3 is a block diagram showing a specific configuration example of a dequantizer 130 and a dequantizer 130 thereof.

FIG. 8 is a lossy quantizer 120 using a hierarchical coding method.
3 is a block diagram showing a specific configuration example of FIG.

9 is a block diagram showing a specific configuration example of an inverse quantization unit 130 corresponding to the lossy quantization unit 120 shown in FIG.

FIG. 10 is an explanatory diagram showing an experimental result of the second embodiment.

FIG. 11 is an explanatory diagram showing experimental results of the second embodiment.

FIG. 12 is a block diagram showing a configuration example of an encoding device and a decoding device according to a third embodiment of the present invention.

FIG. 13 is a block diagram showing a configuration example of an encoding device and a decoding device according to a fourth embodiment of the present invention.

[Explanation of symbols] 110: Frame division section 120 ... Irreversible quantizer 130, 230 ... Inverse quantizer 140 ... Subtraction unit 150, 310 ... Lossless coding unit 160, 220 …… Narakabe part 210, 330 ... Lossless decoding unit 240, 340 ... adder 250 ... Frame composition section 270 ... Selector 280 ... Switching unit 290, 300, 320 ... Bit shift unit

─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Takeshi Mori 2-3-1 Otemachi, Chiyoda-ku, Tokyo Nippon Telegraph and Telephone Corporation (72) Inventor Kazuaki Chikira 2-chome, Otemachi, Chiyoda-ku, Tokyo No. 3 in Nippon Telegraph and Telephone Corporation (56) Reference JP-A-9-252407 (JP, A) JP-A-9-93440 (JP, A) JP-A-8-139935 (JP, A) JP Flat 9-327018 (JP, A) (58) Fields investigated (Int.Cl. ⁷ , DB name) H03M 7/30 G10L 19/00 H04N 7/30

Claims (30)

(57) [Claims] 1. A sample represented by a plurality of code absolute values
Of the digital input signal for each frame
A Ijitaru signal encoding method, and generating a lossy compression code the <br/> Rukoto to lossy encoding said digital input signal for each frame, the inverse quantization corresponding to the lossy coding reversible and generating a local reproduction signal <br/> No., and generating an error signal between the digital input signal and the local reproduction signal by lossless compression-encoding the error signal for each bit position by digital signal encoding method characterized by having a step of outputting the compressed code. 2. A sample represented by a plurality of code absolute values
Of the digital input signal for each frame
A Ijitaru signal encoding method, and generating a lossy compression code the <br/> Rukoto to lossy encoding said digital input signal for each frame, the inverse quantization corresponding to the lossy coding local reproduction and generating a signal <br/> No., and generating an error signal between the digital input signal and the local reproduction signals, and selects and outputs one of said digital input signal and said error signal by And a lossless compression code for each bit position of the selected output signal.
And a step of outputting a lossless compression code by converting the digital signal into a digital signal coding method . 3. Is the sample represented by a plurality of sign absolute values?
Of the digital input signal for each frame
A Ijitaru signal encoding method, upper t of the digital input signal (t is a natural number) and generating a first lossless compression codes by lossless compression encoding of bits, the remaining bits of the digital input signal To the upper direction t
The process of bit arithmetic shift and the non-reversible process by lossy-encoding the digital input signal, which is t-bit arithmetically shifted in the upper direction, for each frame
The process of generating the decompression code and the dequantization corresponding to the lossy coding decompress the local reproduction signal.
And generating an item, the steps of rounding the lower t bits of the absolute value portion of the amplitude of the local reproduction signals, comprising the steps of t bit arithmetic shift local reproduction signals that are rounded to the lower direction, said digital input signal and A step of generating an error signal with the local reproduction signal which is arithmetically shifted in the lower direction by t bits, and a step of outputting a second reversible compression code by lossless compression-encoding the error signal. Is characterized by having
Digital signal encoding method . 4. The process of generating the lossy compression code comprises:
Select one method from among a plurality of types of lossy compression method, a digital signal encoding method according to any one of claims 1 to 3, characterized in that to perform lossy compression on the selected method . 5. The process of generating the lossy compression code comprises:
Digital signal encoding method as claimed in any one of claims 3, characterized in that by encoding by transform coding method the digital input signal is a process of obtaining the lossy compression coding. 6. The process of generating the lossy compression code comprises:
The frequency of the digital input signal is f ₁ , f ₂ , ..., F
_n-1 (f ₁ <f ₂ <... <f _n-1 <f _n ; f _n is the highest frequency of the input signal) is divided into n sections (n is an integer of 2 or more) and encoded. A first band selecting step of selecting a first band signal having a frequency of f ₁ or less from the input signal, and a first code by encoding the first band signal with a first encoding method. The ( _i-1 ) th band decoding that obtains the i-1st decoded signal of frequency f _i-1 or lower from the first encoding process to be generated and each code of 1-i or lower (i = 2, 3, ..., N) And an i-th band selection step of selecting an i-th band signal having a frequency of f _i or less from the input signal, and an i-th difference signal by subtracting the i-1 th decoded signal from the i-th band signal. Obtaining the i-th difference process, and generating the i-th code by encoding the i-th difference signal using an i-th encoding method. An i-th encoding step, and a step of obtaining the quantized signal by obtaining an n-th decoded signal having a maximum frequency f _n or less from each of the n-th or less codes (i = 2, 3, ..., N). digital signal marks according to any one of claims 1 to 3, characterized in that it comprises a
Encoding method . 7. The i-th encoding method (i = 1, 2, ...,
7. The digital signal coding method according to claim 6 , wherein each of n) is a transform coding method . 8. A digital signal decoding for decoding a digital signal by inputting a lossy compression code and a lossless compression code.
A method of generating an error signal for each bit position by decoding the lossless compression code for each frame, and a local reproduction signal by decoding the lossy compression code.
A method for decoding a digital signal , comprising: a step of generating a signal; and a step of reproducing the digital signal by adding the error signal and the local reproduction signal. 9. A lossy compression code, a lossless compression code and a signal
Di for decoding digital signal selection information is input
A digital signal decoding method , which includes a step of generating a decoded signal for each bit position by decoding the lossless compression code for each frame, and a local reproduction signal by decoding the lossy compression code.
And generating an item, comprising the steps of reproducing <br/> said digital signal by adding the local reproduction <br/> signal and the decoded signal based on the signal selection information, directly outputs the decoded signal in particular
The digital signal is reproduced.
Digitally, characterized by having a step of run-option
Signal decoding method . 10. A lossy compression coding the digital signal decoding method of the first lossless compression encoding <br/> second lossless compression coding to decode is inputted digital signal, the first and generating an error signal by decoding each frame lossless compression encoding of the first decoded by decoding said lossy compression code
A step of generating a signal, a step of generating a local reproduction signal by arithmetically shifting the decoded signal in the lower direction by t (t is a natural number) bits, a step of adding the error signal and the local reproduction signal , A step of decoding the second lossless compression code for each frame, and a second decoded signal obtained by decoding the second lossless compression code ,
Di, characterized in that it comprises a step of connecting the upper bit side of said error signal and said local reproduced signal and the addition result
Digital signal decoding method . 11. The process of decoding the lossy compression code into a local reproduction signal is characterized by selecting one method from a plurality of types of lossy compression decoding methods and decoding by the selected method. The digital signal decoding method according to any one of claims 8 to 10 . 12. The process of decoding the lossy compression code into a local reproduction signal is a process of obtaining the local reproduction signal by decoding the lossy compression code by a conversion decoding method. The digital signal decoding method according to any one of claims 8 to 10 . 13. The step of decoding the lossy compression code into a local reproduction signal includes the steps of separating the lossy compression code into first to m-th codes (m is a natural number of 2 or more), and the i-th code. By decoding (i = 1, 2, ..., M) by the i-th decoding method, the i- _th frequency whose frequency is f _i or less
And the i decoding process to generate a decoded signal, from claim 8, characterized in that it comprises a step of generating the local reproduction signal by adding the first decoded signal to the m decoding signals according to claim 10 Described in either
Digital signal decoding method . 14. The i-th decoding method (i = 1, 2,
..., m) is a digital signal decoding method of claim 13, wherein the both a transformation and decoding method. 15. A sample represented by a plurality of code absolute values
Encode the digital input signal for each frame of
A digital signal encoding apparatus, comprising: means for generating a lossy compression code the <br/> Rukoto to lossy encoding said digital input signal for each frame, the inverse quantization corresponding to the lossy coding reversible means for generating a local reproduction signal <br/> No., means for generating an error signal between the digital input signal and the local reproduction signal by lossless compression-encoding the error signal for each bit position by A digital signal encoding device, comprising: means for outputting a compression code . 16. A sample represented by a plurality of sign absolute values
Encode the digital input signal for each frame of
A digital signal encoding apparatus, comprising: means for generating a lossy compression code the <br/> Rukoto to lossy encoding said digital input signal for each frame, the inverse quantization corresponding to the lossy coding local reproduction means for generating a signal <br/> No., means for generating an error signal between the digital input signal and the local reproduction signals, and selects and outputs one of said digital input signal and said error signal by And a lossless compression code for each bit position of the selectively output signal.
And a means for outputting a reversible compression code by converting the digital signal into a digital signal coding apparatus . 17. A sample represented by a plurality of sign absolute values
Encode the digital input signal for each frame of
A digital signal encoding apparatus, upper t of the digital input signal (t is a natural number) means for generating a first lossless compression codes by lossless compression encoding of bits, the remaining bits of the digital input signal To the upper direction t
Bit arithmetic shift means and irreversible encoding by irreversibly encoding the digital input signal that has been arithmetically shifted by t bits in the upper direction
A means for generating an inverse compression code and a local reproduction signal by means of inverse quantization corresponding to the lossy encoding.
Means for generating a No., and means for rounding a lower-order t bits of the absolute value portion of the amplitude of the local reproduction signals, and means for t-bit arithmetic shift local reproduction signals that are rounded to the lower direction, said digital input signal and Means for generating an error signal with the local reproduction signal that has been arithmetically shifted in the lower direction by t bits, and means for outputting a second reversible compression code by lossless compression-encoding the error signal. It is characterized by having
Digital signal encoder . 18. The lossy compression code generation means selects one of a plurality of types of lossy compression methods and performs lossy compression by the selected method. di according to any one of claims 17 to 15
Digital signal encoder . 19. means for generating the lossy compressed code from claim 15, characterized in that the means for obtaining the lossy compression code by encoding the transform coding method the digital input signal The digital signal encoding device according to claim 17 . 20. The means for generating the irreversible compression code processes the digital input signal with frequencies f ₁ , f ₂ ,.
f _n-1 (f ₁ <f ₂ <... <f _n-1 <f _n ; f _n is the highest frequency of the input signal) is divided into n sections (n is an integer of 2 or more) and encoded. Means for selecting a first band signal having a frequency of f ₁ or less from the input signal, and a first code by encoding the first band signal by a first encoding method. And an i-1th band for obtaining an i-1th decoded signal having a frequency of f _i-1 or less from each of the 1-i and subsequent codes (i = 2, 3, ..., N). decoding means, and the i band selection means for frequency from said input signal is selected following the i band signal f _i, the i-difference signal by subtracting the first i-1 decoded signal from the i-th band signal An i-th difference means for obtaining an i-th code by encoding the i-th difference signal by an i-th encoding method And the i encoding means for generating said first n following the code (i = 2,3, ..., n) means for obtaining said quantized signal by obtaining the following n-th decoded signal up to a frequency f _n from The digital signal according to any one of claims 15 to 17 , further comprising:
No. encoder . 21. The i-th encoding method (i = 1, 2,
21. The digital signal coding device according to claim 20 , wherein each of the ..., N) is a transform coding method. 22. the lossy compression coding and lossless compression codes are input digital signal recovery for decoding a digital signal
An encoding device for generating an error signal for each bit position by decoding the lossless compression code for each frame; and a local reproduction signal for decoding the lossy compression code.
Digital signal decoding apparatus , further comprising means for generating a signal and means for reproducing the digital signal by adding the error signal and the local reproduction signal.
Place 23. A lossy compression code, a lossless compression code and a signal.
De for decoding digital signal No. selection information is input
An apparatus for decoding a digital signal , comprising means for generating a decoded signal for each bit position by decoding the lossless compression code for each frame, and a local reproduction signal for decoding the lossy compression code.
Means for generating a No., and means for reproducing <br/> said digital signal by adding the local reproduction <br/> signal and the decoded signal based on the signal selection information, directly outputs the decoded signal in particular
Or a means for reproducing the digital signal.
Digitally, characterized by comprising a means for performing in-option
Signal decoding device . 24. A lossy compression coding the digital signal decoding apparatus of the first lossless compression encoding <br/> second lossless compression coding to decode is inputted digital signal, the first means for generating an error signal by decoding each frame lossless compression encoding of the first decoded by decoding said lossy compression code
A means for generating a signal, a means for generating a local reproduction signal by arithmetically shifting the decoded signal in the lower direction by t (t is a natural number) bits, a means for adding the error signal and the local reproduction signal , Means for decoding the second lossless compression code for each frame, and a second decoded signal obtained by decoding the second lossless compression code ,
Di characterized in that a means for connecting to the upper bit side of said error signal and said local reproduced signal and the addition result
Digital signal decoding device . 25. The means for decoding the irreversible compression code into a locally reproduced signal selects one method from a plurality of kinds of irreversible compression decoding methods, and decodes by the selected method. The digital signal decoding device according to any one of claims 22 to 24 . 26. means for decoding the local reproduction signals the lossy compression code, billing, which is a means for obtaining the local reproduction signal by decoding by the lossy compression coding the transform decoding method The digital signal decoding device according to any one of claims 22 to 24 . 27. A means for decoding the lossy compression code into a local reproduction signal , a means for separating the lossy compression code into first to m-th codes (m is a natural number of 2 or more), and the i-th code. By decoding (i = 1, 2, ..., M) by the i-th decoding method, the i- _th frequency whose frequency is f _i or less
25. An i-th decoding means for generating a decoded signal, and a means for generating the local reproduction signal by adding the first to m-th decoded signals, 25 to 24. 5. The digital signal decoding device according to any one of 1. 28. The i-th decoding method (i = 1, 2,
28. The digital signal decoding apparatus according to claim 27 , wherein each of the ..., M) is a conversion decoding method. 29. The daisy according to any one of claims 1 to 7
A recording medium recording a program for causing a computer to execute each step of the digital signal encoding method . 30. Di according to any one of claims 8 to 14
A recording medium recording a program for causing a computer to execute each step of the digital signal decoding method .