JP3164808B2 - Variable frame vocoder - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a variable-length frame vocoder.

[Conventional technology]

Instead of sending out all the spectral envelope parameters obtained by analyzing the input speech signal for each analysis frame of a certain length in accordance with the analysis frame, N (for each 20 (for example, 20) analysis frames) For example, by selecting 5) representative analysis frames and transmitting the number of analysis frames represented by these representative analysis frames, the so-called repeat bit, together with the spectral envelope parameter of each representative analysis frame. 2. Description of the Related Art Variable-length frame vocoders that reduce the amount of transmission information are known.

The representative analysis frame selected for each section is obtained by approximating the distribution of the spectral envelope parameters indicated by all the analysis frames for each section with an optimal approximation function, and this optimal approximation function is an approximate function such as a rectangle, a trapezoid, or a straight line. Are used for different purposes, and are usually called Dynamic Programming (D
Function settings have been made via P). An example of the piecewise optimal function approximation through the DP method is described in detail in JP-A-62-998.

[Problems to be solved by the invention]

Distortion in the variable-length frame type vocoder is caused by the representation of the representative analysis frame itself due to the finite number of analysis orders and quantization, and the substitution caused by substituting each analysis frame with the representative analysis frame. There is distortion. The nature of speech is time-varying, especially if the order of analysis, the number of quantizations, the number of representative analysis frames, etc. are constant due to the time-varying nature of the spectral envelope, the proportion of each of these distortions in the total distortion changes every moment I do.

However, the above-mentioned conventional variable-length frame vocoder has a drawback that optimization from the viewpoint of total distortion cannot be expected in most of the sections because the number of representative analysis frames in the section is fixed.

An object of the present invention is to provide a variable-length frame type vocoder that can be more optimized than before in terms of total distortion.

[Means for solving the problem]

A variable-length frame vocoder according to the present invention includes means for optimizing the number of representative analysis frames in a section within a predetermined range in a variable-length frame vocoder using piecewise optimal function approximation.

〔Example〕

 Next, the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing the analysis side in the first embodiment of the present invention, and FIG. 2 is a block diagram showing the synthesis side in the same manner.

The analysis side shown in FIG. 1 includes an A / D converter 1 and an LPC analyzer 2
, Sound source analyzer 3, quantizer 4, frame selector 5
7, a representative frame number determiner 8, and a multiplexing circuit 9.

The frame selector 5 includes a quantizer 51, a decoder 52,
It comprises a distortion calculator 53, buffer memories 54 to 57, and a piecewise optimum approximator 58. The frame selectors 6 and 7 are also configured in the same manner as the frame selector 5 except that the numerical conditions of the processes performed by the respective components are not necessarily equal.

An input audio signal whose band is limited by a low-pass filter having a cutoff frequency of 3 and 4 kHz is sampled by the A / D converter 1 at a sampling frequency of 8 kHz and quantized to a predetermined number of bits.

The LPC analyzer 2 performs window processing and analysis on the quantized audio signal input from the A / D converter 1. The quantized audio signal is read out at a period of 10 ms while being stored in the built-in memory at intervals of 30 ms, subjected to window processing by the Hamming function, and output as a 10 ms analysis frame. The quantized speech signal for each analysis frame is subjected to linear prediction analysis, and in this embodiment, the linear prediction coefficient (Li
An eighth-order K parameter, which is one of the near prediction coefficients (LPC), is extracted.

The frame selector 5 divides 20 consecutive analysis frames, that is, 200 ms, into one section, sets the number of representative analysis frames in the section to 4, and divides the quantization distortion and the frame substitution distortion so that the total amount thereof is minimized. Approximate optimal function.

The quantizer 51 quantizes the K parameters input from the LPC analyzer 2 for each analysis frame by the number of quantizations that can be taken when the number of representative analysis frames is 4, and buffers 20 frames of the quantized K parameters. Write to memory 54.
The decoder 52 decodes the quantized K parameter input from the quantizer 51 and writes the decoded K parameter for 20 frames into the buffer memory 55. The distortion calculator 53 is an LPC
The K-parameter input from the analyzer 2 and the K-parameter input from the decoder 52 are used to calculate the quantization distortion for each analysis frame, and the calculated quantization distortion for 20 frames is written into the buffer memory 56. 20 frames of K parameters output from the LPC analyzer 2 before quantization are also stored in the buffer memory.
Written in 57. The piecewise optimal approximator 58 uses the stored contents of the buffer memories 54 to 57 to perform a piecewise optimal function approximation via the DP method so that the total amount of the quantization distortion and the frame substitution distortion is minimized. The representative analysis frame is extracted, the K parameter quantized of each extracted representative analysis frame is output as a quantization parameter 102, the number of analysis frames represented by each representative analysis frame is output as a repeat bit 103, The total amount of quantization distortion and frame substitution distortion of each representative analysis frame is output as total distortion 101.

Incidentally, the piecewise optimal approximator 58 is disclosed in Japanese Patent Laid-Open No. 62-99
Since the operation of the frame selector 15 described in Japanese Patent Publication No. 8 (in the case where quantization is performed by pattern matching) is generalized, the processing algorithm may be used. Is easy to realize.

The frame selector 6 sets the number of representative analysis frames to 5,
The operation is the same as that of the frame selector 5 except that the number of quantizations of the K parameter is reduced accordingly. The frame selector 7 operates in the same manner as the frame analyzer 5 except that the number of representative analysis frames is 6, and the number of quantizations is further reduced accordingly.

The representative frame number determiner 8 compares the total distortions input from the frame selectors 5 to 7, selects the quantization parameter and the repeat bit input from the frame selector having the minimum total distortion, and selects the quantization parameters 202 and Repeat bit 203
Output as Also, the number of representative analysis frames of the selected frame selector is output as the number of representative frames 204.

The sound source analyzer 3 performs window processing on the quantized audio signal input from the A / D converter 1 with a rectangular function at an analysis frame period of 10 ms, and performs a voiced / unvoiced discrimination and pitch period analysis for each analysis frame by a known method. Extract data. These data are quantized by the quantizer 4 to become sound source data 205.

The multiplexing circuit 9 multiplexes the input quantization parameter 202, repeat bit 203, representative frame number 204, and sound source data 205 in a predetermined format determined by the representative frame number 204, and transmits the multiplexed data to the combining side.

The above is the overall description of the operation on the analysis side shown in FIG. Next, the operation of the main components on the analysis side shown in FIG. 1 when the present invention is applied to a 1200 BPS vocoder will be described in detail.

FIGS. 5A, 5B, and 5C show the multiplexing circuit 9 when 4, 5, and 6 are selected as the number N of representative analysis frames in a section having M = 20. 3 shows a frame configuration of output data for one section. As shown in FIG. 5 (a), N =
In the case of 35, 35 bits are assigned to each of the K parameters of the first to fourth representative frames. ₈ , 6, ₅ , 4, ₃ , ₃ , ₃ , and 3 bits are assigned to the K parameters k ₁ , k ₂ ,..., K ₈ of each representative frame, respectively. 20 sets of K parameter vectors analyzed every 10mSEC by LPC analyzer 2 Are each quantized to a desired number of bits, and a quantized K parameter vector Is sent to the buffer memory 54 and the decoder 52.
The buffer memory 54 stores the quantization K parameter vector. The decryptor 52 decrypts the quantized K parameter vector, and converts this to the decrypted K parameter vector. To the buffer memory 55 and the distortion calculator 53. The buffer memory 55 stores the decrypted K parameter vector. The distortion calculator 53 calculates the quantization distortion δ ^(j) , (j = 1,..., 20) for 20 frames by the following equation (1).

Here, w _i (i = 1,..., 8) is the spectral sensitivity of the K parameter. The calculated 20 quantization distortions δ ^(j) , (j = 1,
.., 20) are written to the buffer memory 56.

The piecewise optimal approximator 58 is a K-parameter vector written in the buffer memory 57. And the decrypted K parameter vector written in the buffer memory 55 And the quantization distortion δ ^(j) written in the buffer memory 56,
(J = 1, ..., 20 ) with the total distortion ▲ D of the smallest possible in the case of using the algorithm disclosed in JP-A-62-998 representative frame number N = ^{4 (4)} _The representative frames from which _min ▼ can be obtained and the sections represented by these representative frames are determined. In this embodiment, the maximum section length represented by each representative frame is limited to 16 frames. This limiting process is performed to prevent unnecessary assignment of repeat bits, and this limit has almost no effect on sound quality. If the maximum section length is not limited, the maximum section length that can be taken is 17. Total strain D ⁽⁴⁾
_min ▼ is output via the output line 101, the quantization K parameter of each representative frame is read from the buffer memory 54, and the section length represented by each representative frame is output via the output line 102 as a repeat bit. Each of them is output to the representative frame number determiner 8. The representative frame number determiner 8 Becomes ▲ D ⁽⁾ _min ▼ is searched, and the optimum number of representative frames is determined based on the search result. ▲ D ⁽⁵⁾ _min ▼ 、 ▲
D ⁽⁶⁾ _min ▼ is the total distortion calculated by the frame selector 6 and the frame selector 7, respectively. When = 4 is optimal, the quantization K parameter and repeat bit determined and quantized and selected by the frame selector 5 are output to the multiplexing circuit 9.
When = 4, the multiplexing circuit 9 assembles the frame having the configuration shown in FIG. 5A and outputs the frame to the combining side. The configuration of the sound source parameters is fixed irrespective of the parameters, and is the same as that in FIG. 5 (C). Of course, M = 20 is fixed, and if any three repeat bits are transmitted, there is no need to transmit the other one.

The frame selector 6 controls the processing in the case of = 5, and the maximum section length is set to 9. The frame selector 7 controls the processing in the case of = 6, and the maximum section length is set to 9. FIG. 5 (A),
In (B) and (C), the first bit "F" is a bit for frame synchronization, and the second and third bits "00" and "01" in FIGS. 5A and 5B, and Second bit "1" in FIG. 5 (C)
Is a bit that specifies the number of representative frames.

 Next, the overall operation of the synthesizing side will be described with reference to the drawings.

The combining side shown in FIG. 2 includes a separating circuit 11 and a demultiplexer 12.
, A parameter interpolator 13, a buffer memory 14, a decoder 15, a pulse generator 16, a noise generator 17, and a switch
18, an LPC synthesis filter 19, and a D / A converter 20.

The separation circuit 11 separates the multiplexed signal transmitted from the multiplexing circuit 9 and sends the quantization parameter 202 to the demultiplexer 12.
Repeat bit 203 to parameter interpolator 13, representative frame number 204 to decryptor 12 and parameter interpolator 13, and
The sound source data 205 is output to the decoder 15.

The decryptor 12 decrypts the quantization parameter 202 with reference to the representative frame number 204, converts the decrypted K parameter into an α parameter, and outputs it to the parameter interpolator 13. The parameter interpolator 13 refers to the representative frame number 204, and repeats the input α parameter as many times as the number of analysis frames specified by the repeat bit 203, and writes the same into the buffer memory 14.

The decoder 15 decodes the sound source data 205, outputs pitch cycle data to the pulse generator 16, and controls the switch 18 by voiced / unvoiced discrimination. When voiced, the pulse generator 16 generates a pulse train at the pitch cycle specified by the input pitch cycle data.
Is selected and output. When there is no voice, the white noise output from the noise generator 17 is selectively output by the switch 18.

The LPC synthesis filter 19 is driven by a sound source input from the switch 18 and synthesizes a digital audio signal using the α parameter read from the buffer memory 14 for each analysis frame as a filter coefficient. This digital audio signal is converted into an analog signal by the D / A converter 20 to be an outgoing / transmitting audio signal.

As described above, in the embodiment shown in FIGS. 1 and 2, the number of representative frames in a section is set within the range of 4 to 6 (along with the quantization number of the K parameter), and the quantization distortion and the frame replacement distortion are reduced. And the adaptive control is performed so that the total amount becomes minimum.

The embodiment shown in FIG. 1 and FIG.
Although a parameter is used, other LPCs such as an α parameter and a line spectrum pair (LSP) can also be used.

The embodiment shown in FIGS. 1 and 2 has been described above.

FIG. 3 is a block diagram showing the analysis side in the second embodiment of the present invention, and FIG. 4 is a block diagram showing the synthesis side in the same manner.

The analysis side shown in FIG. 3 includes the analysis side LPC analyzer 2, the frame selectors 5 to 7, the representative frame number determiner 8, and the multiplexing circuit 9 which are the LPC analyzer 31, the frame selector 32 shown in FIG. , An LPC order determiner 35, and a multiplexing circuit 36.
The frame selectors 32 to 34 have the same configuration as the frame selector 5 on the analysis side shown in FIG. 1 except that the numerical conditions of the processing performed by the respective components are not always equal. The components are assigned reference numerals 61 to 68 so as to correspond to the reference numerals 51 to 58 assigned to the respective components of the frame selector 5. The combining side shown in FIG. 4 includes a combining circuit 11, a decoding unit 12, a parameter interpolator 13, a buffer memory 14, and an LPC combining filter 19 on the combining side shown in FIG.
1, Decoder 42, Parameter interpolator 43, Buffer memory 44, L
It is configured by replacing with a PC synthesis filter 45.

The operation of the LPC analyzer 31 is the same as the operation of the LPC analyzer 2 except that the analysis order is twelfth.

The frame selector 32 treats the 12th-order K parameter input from the LPC analyzer 31 as an 8th-order K parameter, and
The representative analysis frame number in the section of s is set to 6, and the piecewise optimal function approximation is performed so that the total amount of the quantization distortion and the frame substitution distortion is minimized.

The quantizer 61 ignores the ninth and subsequent K parameters of the 12th K parameter input for each analysis frame from the LPC
It is treated as the next K parameter and quantization is performed. Strain calculator 63
Calculates the distortion of quantization in the quantizer 61 on the basis of the unquantized 12-order K parameter. The piecewise optimal approximator 68 uses the distortion generated when substituting the unquantized 12th order K parameter with the 8th order decoding K parameter (of the representative analysis frame) as the frame substitution distortion. Except as described above, the operation of the frame selector 32 is the same as the operation of the frame selector 5.

The frame analyzer 33 calculates the 12th order K parameter to be input.
The operation is the same as that of the frame selector 32 except that the number of representative analysis frames is set to 5 by treating it as a 10th-order K parameter. The frame selector 34 also treats the input 12th-order K parameter as it is as 12th order, and reduces the number of representative analysis frames to 4
The operation is the same as that of the frame selector 32 except that

The LPC order determiner 35 converts the order of the K parameter handled by the selected frame selector in place of the representative frame number 204 into an LPC order.
The operation is the same as that of the representative frame number determiner 8 except that it is output as the degree 206.

Multiplexing circuit 36, demultiplexing circuit 41, demultiplexer 42, parameter interpolator 43, phase of buffer memory 44 with multiplexing circuit 9, demultiplexing circuit 11, demultiplexer 12, parameter interpolator 13, buffer memory 14 The differences are due to changes based on differences in the format of the parameters handled by these. LPC synthesis filter 19 is 8
The LPC synthesis filter 45 is configured as a twelfth-order all-pole filter, while being configured as the next all-pole filter.

As can be seen from the differences described above with respect to the embodiment shown in FIGS. 1 and 2, the embodiment shown in FIGS.
The number of representative analysis frames in the section along with the order of C is 4 to 6
Adaptive control is performed in a range of

The embodiment shown in FIG. 3 and FIG.
Parameters are used, but other LPCs also use the 12th order L
A frame selector that converts a PC into an 8th or 10th order LPC
It can be used if provided at the input terminals of 32 and 33.

The embodiment shown in FIGS. 3 and 4 has been described above.

Although the two embodiments described above optimize the number of representative analysis frames in a section and either the number of quantizations or the LPC order, the number of representative analysis frames, the number of quantizations, and the LP
It is also possible to perform the optimization in consideration of all the C orders.

〔The invention's effect〕

As described above, the variable-length frame vocoder of the present invention can optimize from the viewpoint of total distortion by adaptively controlling the number of representative analysis frames in a section, and has the effect of reducing total distortion. .

[Brief description of the drawings]

FIG. 1 is a block diagram showing an analysis side in the first embodiment of the present invention, FIG. 2 is a block diagram showing the synthesis side similarly, and FIG. 3 is a block diagram showing an analysis side in the second embodiment of the present invention. FIG. 4 is a block diagram showing the combining side, and FIG.
The figure shows the structure of the frame. 1 ... A / D converter, 2,31 ... LPC analyzer, 3 ... Sound source analyzer, 4 ... Quantizer, 5-7,32-34 ... Frame analyzer, 8 ... Representative frame Number determiner, 9,36 …… Multiplexer, 11,41 …… Separator, 12,42 …… Decoder, 13,43…
... parameter interpolator, 14,44 ... buffer memory, 15 ...
... Decoder, 16 ... Pulse generator, 17 ... Noise generator,
18… Switch, 19,45… LPC synthesis filter, 20… D / A
Transformer, 35 …… LPC order determiner, 51,61 …… Quantizer, 5
2,62 …… Decoder, 53,63 …… Strain calculator, 54-57,64-67
…… Buffer memory, 58,68 …… Piecewise optimal approximator.

Claims (3)

(57) [Claims] 1. A variable length frame type vocoder using piecewise optimal function approximation, an analyzer for analyzing an input speech signal at predetermined intervals to extract a characteristic parameter, and receiving an output of the analyzer for quantization. A plurality of frame selectors that perform a piecewise optimal function approximation so that the total distortion that is the sum of the distortion and the frame substitution distortion is minimized, wherein the number of representative analysis frames in the division is different from each other. A variable length frame type vocoder comprising: a frame selector; and a representative frame number selector which receives outputs of the plurality of frame selectors and selects an output of the frame selector having the smallest total distortion. 2. The variable-length frame vocoder according to claim 1, wherein said plurality of frame selectors perform quantization with different numbers of quantization bits according to the number of representative analysis frames. 3. A variable-length frame-type vocoder using piecewise optimal function approximation, wherein an analyzer extracts an characteristic parameter by LPC-analyzing an input speech signal at a predetermined analysis order every predetermined period. A plurality of frame selectors that receive the output and perform a piecewise optimal function approximation so that the total distortion, which is the sum of the quantization distortion and the frame substitution distortion, is minimized. The plurality of frame selectors that perform the piecewise optimal function approximation with different representative analysis frames according to the analysis order to be considered while differentiating the analysis orders, and receiving the outputs of the plurality of frame selectors, A variable-length frame type vocoder comprising: an LPC order selector for selecting an output of a frame selector having the smallest distortion.