JPH02143712A - Digital signal processor - Google Patents

Abstract

PURPOSE:To reduce a quantizing error by selecting a proper quantized value and a threshold level for each prescribed period so as to quantize the input signal. CONSTITUTION:The unit is provided with an optimum quantizing analyzer 6 selecting a quantized value and a threshold level based on an input signal S21 for each prescribed period and quantizers 7,8,10 - 12 quantizing the input signal S21 based on the quantized value and the threshold level. Then the quantized value and the threshold level are selected for each prescribed period based on the input signal S21 and the input signal S21 is quantized based on the quantized value and the threshold level. Thus, the quantizing error is suppressed.

Description

DETAILED DESCRIPTION OF THE INVENTION A. Field of Industrial Application The present invention relates to a digital signal processing device, and is suitable for use in, for example, a digital signal processing device designed to record, reproduce, and transmit audio signals and the like in high quality. It is something.

3. Detailed description of the invention The present invention will be explained in the following order.

A: Industrial field of application B: Overview of the invention C: Prior art (Fig. 3) D: Problem to be solved by the invention (Fig. 3) E: Means for solving the problem (Fig. 1) Overview The present invention makes it possible to reduce quantization errors in a digital signal processing device by selecting an optimal quantization value and threshold value and quantizing an input signal every predetermined period.

C. Prior Art Conventionally, in this type of digital signal processing device, an adaptive predictive coding method has been used.
vecoding: By encoding and transmitting the audio signal using the APC method, the S/N
There is a method that prevents deterioration of ratio, clarity, etc. and transmits data with high transmission efficiency (Japanese Patent Laid-Open No. 59-223
033, JP 60-223034, JP 61-158217, JP 61-15821
Publication No. 8, Japanese Patent Application No. 63-46595, Japanese Patent Application No. 63-63
No. 1724, Patent Application No. 63-61725, Patent Application No. 63-
No. 65192).

That is, an input digital signal is received by a predetermined prediction filter, and a residual signal consisting of a difference signal between the output signal of the prediction filter and the input digital signal is obtained.

At this time, the input digital signal is divided into blocks for each predetermined period, and each block is subjected to linear predictive analysis (linear predictive analysis).
Predictive coding: LP
By applying the method C) and switching the frequency characteristics of the prediction filter, the residual signal is made small.

Further, as shown in FIG. 3, each block BLI, BL2
After bit-shifting the residual signal SKI according to the maximum value of the residual signal 521 (FIG. 3(A)) of , ..., for example, 2-bit data is transmitted from the most significant bit. .

Therefore, for each block, the quantized values y0, yl, )'z,
yi (FIG. 3(B)) is switched, thereby switching the dynamic range of the data to be transmitted and requantizing the residual signal S2I.

Thus, by requantizing the residual signal and transmitting the requantized residual signal together with the coefficients of the prediction filter (that is, representing the frequency characteristics of the prediction filter) and the bit shift amount, the input digital signal is Compared to the case of direct transmission, input digital signals can be transmitted with higher transmission efficiency, and deterioration of S/N ratio, clarity, etc. can be prevented.

Problem to be solved by the invention D By the way, in this type of digital signal processing device, it is inevitable that errors occur during requantization, and therefore, when decoding on the transmission side, this error becomes noise. It is inevitable that they will be superimposed as components.

Therefore, if this noise component (hereinafter referred to as requantization error) can be suppressed, it is considered possible to transmit a desired input digital signal with even higher quality.

The present invention has been made in consideration of the above points, and it is an object of the present invention to propose a digital signal processing device that is capable of suppressing requantization errors compared to the prior art.

Means for Solving the EI'J1 Problem In order to solve this problem, the present invention uses optimal quantization analysis to select the quantization value yj and the threshold value Xj at each predetermined period based on the input signal 321. 6 and quantization (based on IIyj and threshold Xj), the input signal SZ
A quantizer 7.8.10.11 for quantizing I, and a quantizer 2 are provided.

Based on the F-effect input signal SZt, quantization (l
IIyj and threshold value Xj are selected, and the quantized value y,
If the human input signal Sol is quantized based on the threshold value Xj and the threshold value Xj, the quantization error can be suppressed.

Example G (G1) Principle of optimal quantization The requantization error ε is expressed as: x1 is each sample value of the input signal to be requantized; P(x) is the probability distribution of the sample value X; and P(x) is the threshold for requantization. Letting Xj and the requantization value to be yj, the following formula % Formula % Here, M represents the number of requantization values yJ.

Therefore, by differentiating equation (1) with respect to y and X, respectively, we obtain 0
Then, the following formula 'Ij +'li◆1 Z (j=1.2, ......, M-1) (j=1.2,
. . . , M) . . . The relationship (3) is obtained.

Therefore, if the requantization threshold value X and the requantization value yj are selected so as to satisfy the relationships of equations (2) and (3) for each block, the requantization error ε can be minimized. It is possible to requantize it so that it becomes a value (hereinafter referred to as optimal quantization).

(G2) First Embodiment In FIG. 1, 1 indicates a digital signal processing device as a whole, and based on the above-mentioned principle of optimal quantization, the requantization threshold value XJ and the requantization threshold are set for each block. value y,
Select.

That is, the digital signal processing device 1 provides a 16-bit input digital audio signal S to the prediction filter 2, and generates a difference signal between the output signal of the prediction filter 2 and the input digital audio signal S1 via the subtracter 3. A residual signal SZI is obtained.

On the other hand, the LPG analysis circuit 4 divides the input digital audio signal SI into blocks for each section of 20 Ctasec, applies a linear predictive analysis method to each block, and performs prediction that switches the frequency characteristics of the prediction filter 2. output filter parameter ptz.

Thereby, the frequency characteristics of the prediction filter 2 are switched so that the maximum value of the residual signal S2I becomes the minimum for each block.

On the other hand, the optimal quantization analyzer 6 analyzes the Lloyd max (LI) shown in FIG. 2 for each block of the residual signal SKI.
Execute the processing procedure applying the OYD-MAX) method,
Thereby, the threshold value X and the requantization value yj for requantizing the 16-bit residual signal S2I to 2 bits are selected.

That is, the optimal quantization analyzer 6 determines the requantization threshold x
After initializing i (j=o, 2's internal counter j is set to the value l.

Here, the first internal counter is used to count the number of repetitions of the processing procedure, and the second internal counter j is used to count the number of requantization thresholds X. It will be done.

Next, the optimal quantization analyzer 6 moves to step SP4 and sets the first internal register Y to the value 0, and sets the third internal counter C to the value 0, and then moves to step SP5 to set the first internal register Y to the value 0. 4's internal counter i is set to the value 0.

Here, the fourth internal counter i is used to count the number of samples in l blocks constituting the residual signal S2I.

Next, the optimal quantization analyzer 6 moves to step SP6, and the sample value X(1) (i=0 in this case) of the input residual signal S2I is changed from the threshold value XJ-1 to the threshold value. It is determined whether or not the value is located between the values X70 (in this case, it is in the range from threshold Once obtained, the process moves to step SP7, where the fourth internal counter i is incremented by the value l.

On the other hand, if a positive result is obtained in step SP6, the process moves to step SP8, where the sample value X (!) is added to the first internal register Y, and the third internal counter C is incremented by the value 1. Then, the process moves to the next step SP7.

Thus, the third internal counter C calculates the sample value xtt> located between the threshold value Xj-1 and the threshold value X.
Count the number of.

Next, the optimal quantization analyzer 6 moves to step SP9 and determines whether the value of the fourth internal counter i matches the number of samples N in the block. In this case, a negative result is obtained, and the process proceeds to step SP6. Return to

Therefore, in step SP6, the optimal quantization analyzer 6 calculates the threshold value x0 for the sample value X., (i=1) following the sample value
It is determined whether or not it is located within the range of the threshold value x1.

If a positive result is obtained here, it is added to the first internal register Y and the third internal counter C is incremented in step SP8, whereas if a negative result is obtained, the process moves to step SP9 via step SP7. .

Thus, sequential steps 5P6-SP? -5P9-3P
6 or step 5P6-3P8-3P? -3P9-3P
By repeating the processing loop LOOP 1 of 6, the threshold value x0 is changed to the threshold value x.

Sample values X, M) located in the range of are sequentially added to the first internal register Y, and the added sample value
. , is stored in the third internal counter C.

Furthermore, each time the processing loop LOOP 1 is repeated, the value of the fourth internal counter i is incremented by 1, so if it is repeated N times in total, the value of the fourth internal counter i is increased in step SP7. Set to the value N.

Therefore, since the number of samples in the block is the value N, all the sample values in the block are X(mu).

When the processing loop LOOP l is repeated for , a positive result is obtained in step SP9, and the optimal quantization analyzer 6 moves to step 5 PIO.

Thus, by repeating the processing loop LOOPI, the arithmetic processing of the following formula (%) (4) is executed.

Subsequently, in step 5 PIO, the optimal quantization analyzer 6 calculates the value of the first internal register Y and the third internal counter C.
Based on the value of , the first requantization value y expressed by the following equation 3 is selected.

Thus, after obtaining the sum of sample values X (41) located in the range from threshold x0 to threshold In the range from the value x0 to the threshold value X, a requantization value yl that minimizes the requantization error ε can be selected.

After storing the first requantization value yl in a predetermined register circuit, the optimal quantization analyzer 6 then proceeds to step 5PII and increments the value of the second internal counter j by the value l.

Next, the process moves to step 5P12, and it is determined whether the value of the second internal counter j matches the number M of requantized values yj (in this case, M=4 since requantization is performed to 2 bits). In this case, a negative result is obtained and the process returns to step SP4.

Here, the optimal quantization analyzer 6 subsequently sets the first internal register Y and the third internal counter C to (i 0), then moves to step SP5 and sets the fourth internal counter i to the value 0. setting, and then moves to step SP6 to judge whether or not the sample value X(!l (i=0) of the input residual signal S2I is located in the range from the threshold value x1 to the threshold value x2. do.

In this way, the processing loop LOOPI is repeated again, whereby the first requantization value y1 is selected between the threshold value x0 and the threshold value x1. ! The sample values X (i) between them are sequentially added and stored in the first internal register Y, and the number of added sample values X, □, is stored in the third internal counter C.

When the processing loop LOOPI is repeated for all sample values Xli) in the block, the optimal quantization analyzer 6 moves to step 5PIO and calculates the following equation y2 "- 〇(x,:5X t+><Xri-
... Select the second requantization value y8 expressed by (6), and thereby the requantization value y2 that minimizes the requantization error ε between the threshold value x1 and the threshold value x2. Select.

Subsequently, the optimal quantization analyzer 6 performs the second
After incrementing the internal counter j of
Proceeding to P12, it is determined whether the value of the second internal counter j matches the number M of requantized values yj.

Thus, steps 5P12-5P4-3P5-LOO
PI-3PIO-3PI 1-3P12 processing loop L
By repeating OOP2, sequentially from threshold X, to threshold x1, from threshold x1 to threshold Xyo, from threshold Xt to threshold x8, from threshold X, to threshold X
requantization value that minimizes the requantization error ε during 1)'J
(j-1, 2, 3, 4) can be selected, and thereby the requantization value y expressed by the general formula (3) 2 can be selected.

On the other hand, the requantization value yj between all thresholds
is selected, the value of the second internal counter j is incremented to the value M in step spH.

Therefore, in this case, a positive result is obtained in step 5P12, and the optimal quantization analyzer 6 moves to step 5P13.

Here, the optimal quantization analyzer 6 sets the second internal counter j to the value 1 again, and then moves to step 5P14 to requantize the values yJ and yj,I (since j=1 in this case). , the requantized value yl, and the y value), a second threshold value x1 expressed by the following relationship is selected.

As a result, the second threshold value x1 expressed by the general formula (2) can be selected between the requantization values y1 and 12, and the requantization value 41! A second threshold X that minimizes requantization error considerations between y + and 'It.

can be selected.

Next, the optimal quantization analyzer 6 moves to step 5PI5 and increments the value of the second internal counter j by the value l, and then moves to step 5P16, where the value of the second internal counter j becomes the requantized value y. , the number M (M=4) - determine whether or not the player loses.

In this case, a negative result is obtained in step 5P16, and the optimal quantization analyzer 6 returns to step 5P14.

Thus, after the second internal counter j is incremented, by repeating the step 5P14, the requantization error ε is set between the requantization value yt and 73, similarly to the case of the second threshold value X. The third threshold value x that minimizes
8 is selected.

Therefore, steps 5P14-SP15-3PI6-5P
By repeating 14 processing loops LOOP3, requantization is performed between requantization value y1 and requantization value y8, between requantization value y2 and requantization value y3, and between requantization value y and requantization value 14, respectively. Threshold value Xj that minimizes the quantization error ε
(j=1.2.3) can be selected.

On the other hand, the threshold value Xj
is selected, the value of the second internal counter j is incremented to the value M in step 5P15, which results in a positive result in step 5P16, and the optimal quantization analyzer 6 moves to step 5PIT.

Here, the optimal quantization analyzer 6 increments the value of the first internal counter by the value l, and then in step 5P18
Then, it is determined whether the value of the first internal counter matches the predetermined value 5 or not.

In this case, a negative result is obtained in step 5P18, and the optimal quantization analyzer 6 returns to step SP3.

Therefore, the optimal quantization analyzer 6 executes the processing procedure of step SP3 again, then repeats the processing loop LOOP2, and instead of using the threshold value initialized by this, the optimal quantization analyzer 6 repeats the processing loop LOOP3 and uses the selected threshold value. The requantization value y that minimizes the requantization error recovery is again selected between the values X7.

Subsequently, when the re-quantization value y, is re-selected, the optimal quantization analyzer 6 executes the processing procedure of step 5P13, and then repeats the processing loop LOOP3, thereby re-selecting the re-quantization value yj. Then, the threshold value X that minimizes the requantization error 8 is reset, and the process moves to step 5P17.

In this way, steps 5P18-3P3-LOOP2-3 are repeated until the value in the first internal counter reaches the predetermined value 5.
The processing loop LOOP4 of P13-LOOP3-3P17-3P1B is repeated, and the requantization value yj and the threshold value Xj are alternately selected so that the requantization error of the entire requantization process is converged. Ru.

On the other hand, when the processing loop LOOP4 is repeated five times, a positive result is obtained in step 5P1B, and the optimal quantization analyzer 6 moves to step 5P19 to end the processing procedure, thereby re-quantizing the optimized The conversion value yJ and the threshold value X are selected.

In practice, if the processing loop LOOP4 is repeated five times, the requantization error of the entire requantization process can be optimized within a necessary and sufficient range, and thus the requantization error is lower than the conventional one. can be reduced.

Further, the optimal quantization analyzer 6 uses the threshold value X optimized in the immediately preceding block instead of the threshold value initialized in the subsequent block, and repeats the processing procedure. .

In this case, in the audio signal, there is a correlation between each block, and the threshold value Xj optimized for the previous block
By using as the initial value, the processing loop LOOP
Optimized requantization value yj and threshold value Xj can be obtained even if the number of repetitions of 4 is reduced to about 5 times.

Therefore, the requantization value Yt and the threshold value Xj that are optimized as a whole can be obtained within a short time, thereby simplifying the requantization process.

Furthermore, the optimal quantization analyzer 6 uses the optimized requantization value y and the threshold value
At the same time, it is output to the inverse requantizer 9 to be transmitted.

The requantizer 7 requantizes the residual signal S2I with the optimized requantization value y and threshold value Xj based on the optimal quantization parameter P0, and the resulting 2-bit requantizer output the conversion signal SL to the transmission target.

On the other hand, the inverse requantizer 8 decodes the requantized signal SL based on the optimal quantization parameter P.

The subtracter 10 obtains a difference signal between the output signal of the inverse requantizer 8 and the residual signal sit, thereby generating a requantization error signal S.
Zt is obtained, and the requantization error signal SZZ is output to the prediction filter 11.

The predictive filter 11 has predictive filter parameters pp
z, the signal is switched to the same frequency characteristic as the prediction filter 2, and after correcting the frequency characteristic of the requantization error signal S2□, it is fed back to the requantizer 7 via the subtracter 12. ing.

As a result, the spectrum shape of the requantization error signal S2□ is corrected to a flat spectrum shape with minimum energy.

On the other hand, in the transmission target, the inverse requantizer 9 decodes the requantized signal SL, thereby decoding the 2-bit requantized signal SL into the residual signal SK+ before requantization. It is output to the prediction filter 15 via the subtracter 14.

The prediction filter 15 has a prediction filter parameter PF.
According to Z, the frequency characteristic is switched to the same as that of the prediction filter 2, and the output signal of the inverse requantizer 9 is fed back to the subtracter 14, thereby converting the decoded residual signal S21 into the digital audio signal StO. Decrypt.

Thus, for the transmission target, the 2-bit requantized signal S
t, and the predictive filter parameter P2□ and the optimal quantization parameter P, for each block.
The 16-bit input digital audio signal S1 can be transmitted by simply transmitting the 16-bit input digital audio signal S1.

Incidentally, in this embodiment, the requantizer 7 and the inverse quantizer 8
, subtracters 10 and 12, and prediction filter 11 constitute a quantizer that quantizes the input residual signal Sol using the optimized requantization value yj and threshold value Xj.

According to the above configuration, the optimized requantization value yj and threshold X are obtained through Lloyd Max's method, and the optimized requantization value yj and threshold X are By requantizing the residual signal Sol, requantization errors can be suppressed for each block, and thus the input digital audio signal S1 can be transmitted with higher quality than before.

(G3) Other embodiments In the embodiments described above, the requantization value yJ and the threshold value X, which are optimized by applying Lloyd Max's method,
Although the present invention is not limited to this, the present invention is not limited to this.
For example, MAX's method, Pantilditch (
PANTER-DITE) method, Smith (SMITH)
) can be widely applied.

In this case, for example, when applying the Max method, from the initial value Xj of the threshold value, solve the following formula % formula (9) to sequentially obtain the requantized value yj and the threshold value XJ, and then solve the following formula When the convergence condition is checked and the convergence condition is satisfied, the following formula is set as %, and by repeating formulas (8) and (9),
An optimized requantization value yJ and threshold value X can be obtained.

Here, δ and Δ are appropriate minute amounts.

Further, in the above-described embodiment, a case was described in which a 16-bit input digital audio signal Sl is requantized into a 2-bit requantized signal and transmitted, but the present invention is not limited to this, and the present invention can be applied to various bit amounts, The present invention can be widely applied to requantizing various input digital signals into requantized signals of desired bits.

Further, in the above-described embodiment, a case has been described in which the requantization value y and the threshold value X are sequentially requantized so that the error converges. A requantization value yj and a threshold value X are selected, and from among these, a requantization value y that gives the smallest requantization error is selected.

and the threshold value Xj may be selected.

Furthermore, in the above embodiment, the LPG is
Regarding the analyzed residual signal SZI, the case has been described in which the requantization value y and the threshold value Xj are selected for each block, but the present invention is not limited to this, and the residual signal can be It is also possible to form blocks and select the requantization value yJ and the threshold value Xj.

Furthermore, in the above embodiment, a case was described in which the residual signal 321 obtained as a result of LPG analysis was requantized, but the present invention is not limited to requantizing the residual signal, but also quantizing various input signals. It can be widely applied when

Furthermore, in the above-described embodiments, the case of transmitting a digital signal has been described, but the present invention is not limited to transmitting a digital signal, but can be widely applied to recording and reproducing a desired digital signal.

H Effects of the Invention As described above, according to the present invention, an optimized requantization value and threshold are selected at predetermined intervals, and an input signal is quantized using the requantization value and threshold. By doing so, it is possible to reduce the quantization error that occurs during quantization compared to the conventional method, and thus it is possible to obtain a digital signal processing device that can quantize a desired input signal with high quality.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing a digital signal processing apparatus according to the present invention, FIG. 2 is a flowchart for explaining its operation, and FIG. 3 is a signal waveform diagram for explaining conventional requantization. 1...Digital signal processing device, 2.11.1
5... Prediction filter, 4... LPC
Analysis circuit, 6... Optimal quantization analyzer, 7...
...Requantizer, 8.9...Inverse requantizer.

Claims (1)

[Claims] An optimal quantization analyzer that selects a quantization value and a threshold at predetermined intervals based on an input signal; 1. A digital signal processing device comprising: a quantizer that performs quantization.