JPH0754439B2 - Method of synchronizing variable length frame type code decoding device - Google Patents

Description

The present invention relates to a synchronization method for a variable-length frame type code decoding device, and in particular to a variable-length frame type code decoding for greatly improving the call delay in the synchronization processing. The present invention relates to a device synchronization method.

[Conventional technology]

The input voice is analyzed by a basic analysis frame (hereinafter, simply referred to as an analysis frame), a spectrum envelope parameter as a voice parameter and sound source information are extracted, coded and sent to a decoding side via a transmission line. A coding / decoding device that synthesizes input speech is well known.

In this case, the spectral envelope parameter indicating the macroscopic features of the input speech spectrum is LPC (Linear Prediction Codin
g, linear predictive analysis) coefficient or its derivation form is used, which is α parameter, K parameter, or
Various coefficients such as LSP (Line Spectrum Pairs) are used. The sound source information indicating the microscopic characteristics of the input speech spectrum is necessary for modeling the pulse frequency corresponding to the pitch frequency and the white noise corresponding to the voiced / unvoiced information and encoding / decoding. Pitch frequency, voiced / unvoiced sound, sound source amplitude, etc. are used. In addition, multi-pulses, residual waveforms, etc. may be encoded / decoded by using the sound source waveform information.

The variable-length frame type coding / decoding apparatus, in the above-described coding / decoding apparatus, performs a plurality of analysis instead of faithfully transmitting the spectrum envelope parameter among the speech parameters to be sent from the coding side to the decoding side in the analysis frame unit. By selecting a representative analysis frame that represents a frame and transmitting the spectrum envelope parameter of this representative analysis frame and the number of analysis frames designated as having the same spectrum envelope parameter as this representative analysis frame, so-called repeat bits. It realizes a significant reduction of the transmission bit rate.

There are various kinds of methods for constructing the variable length frame in such a variable length frame type coding / decoding device, but basically the following contents are used.

That is, an optimum approximation relationship that optimally approximates this segment spectrum envelope is set for each segment composed of a preset number M of analysis frames. As a function used for this piecewise optimal approximation, there is, for example, a rectangular function, and its setting is to divide an approximate rectangular envelope such that the total sum of distances between spatial vectors by spectral envelope parameters for each analysis frame is minimized. It is carried out in the form of finding each. A plurality of i analysis frames forming the approximate rectangle obtained in this manner are used as a representative frame representing the section. The above-mentioned i pieces are arbitrarily set corresponding to the respective spectrum envelopes for each section, and such an optimal approximation itself includes the analysis order and the total between spatial vectors by the spectrum envelope parameter for each analysis frame. The sum of the hit vector distances is used as the evaluation scale, and the combination of analysis frames that minimizes this is DP (Dynamic Programming,
It is also well known that it can be easily obtained by dynamic programming.

Now, the spectrum envelope parameter in the variable length frame format thus obtained is encoded in a predetermined format together with the excitation information and transmitted to the decoding side. In this case, the frame synchronization in the data transmission / reception is decoded with the encoding side. It is necessary between the liquorization side and

As the synchronization format, either independent synchronization or forced synchronization is often used, but in any case, one segment forming a variable length frame is treated as a basic frame in the synchronization processing. .

That is, in independent synchronization, the decoding side accumulates data in units of one partition in order to perform synchronization control in order to remove slips due to the clock frequency difference from the code side. Synchronous recovery is performed through frame update by deletion, and the clock provided from the encoding side to the decoding side is generated even in forced synchronization.
In order to eliminate the effects of failures due to distribution, transmission, etc. 1
At present, the data is stored in units of divisions, and this is also premised on frame synchronization by one division shift.

[Problems to be solved by the invention]

The above-described conventional synchronization method of this type of variable-length frame type code decoding apparatus is a method of accumulating data in units of one segment regardless of independent synchronization or forced synchronization and achieving frame synchronization in units of one synchronization frame. Therefore, there is a problem that a call delay cannot be avoided remarkably as compared with an ordinary coding / decoding device that performs frame synchronization in units of analysis frames. As for this call delay, one division of variable length frame
This is obvious considering that it is not unusual for the size to reach 100 mSEC.

An object of the present invention is to eliminate the above-mentioned drawbacks, and in a synchronization method of a variable length frame type coding / decoding device, sufficiently shorter than one segment, for example, an analysis frame or data corresponding to the number of samples of time slots corresponding thereto is By providing a means to achieve frame synchronization only by storing,
It is an object of the present invention to provide a synchronization method for a variable length frame type code decoding device, which significantly improves the call delay based on the synchronization processing.

[Means for solving problems]

The method of the present invention is a method of synchronizing a variable length frame type coding / decoding device that utilizes independent synchronization or forced synchronization for synchronization with a coding side, and a voiced section designated by a representative spectrum envelope parameter for each variable length frame section. Alternatively, the silent side of each variable length frame section or the time length of the voiced section and the silent section is adaptively controlled according to the degree of synchronization deviation between the encoding side and the decoding side, while the encoding side and the decoding side are being controlled. It is configured to include a synchronization holding means for holding the synchronization.

〔Example〕

 The present invention will now be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of the present invention, which comprises an encoding side 1 and a decoding side 2.

The encoding side 1 further includes an A / D converter 11, a window processor 12, an LSP
Analyzer 13, sound source information analyzer 14, variable length frame selector 15,
The clock generator 16 and the multiplexer 17 are provided.

Further, the decoding side 2 includes a demultiplexer 21, a frame synchronizer 22, a fixed length framing device 23, a memory 24, an adaptive controller 25,
It comprises a correction message generator 26, a clock generation circuit 27, a voice synthesizer 28 and the like.

In the embodiment shown in FIG. 1, the case where independent synchronization is used as the synchronization method is used as an example, and the LSP coefficient is used as the spectrum envelope parameter, but this is the forced synchronization method and other LPC coefficients are used. It does not matter if it is done, and basically it can be carried out in substantially the same manner.

Now, the input voice is supplied to the A / D converter 11, and after removing unnecessary high frequency through LPF (Low Pass Filter) of high frequency cutoff frequency 3.4KHz, it is sampled at 8KHz sampling frequency It is digitized in the quantization step.

Upon receiving the quantized audio signal from the A / D converter 11, the window processor 12 receives one quantized audio signal for a preset one block, for example,
Each 30mSEC (240 samples) is temporarily stored in the memory, and the window processing that multiplies this by a Hamming function of the preset window function and cuts out is performed every 10mSEC analysis frame period. Therefore, the basic analysis frame size, that is, the analysis frame length is 10 mSEC.

The output of the window processor 12 is supplied to the LSP analyzer 13 and the sound source information analyzer 14.

The LSP analyzer 13 performs LPC analysis on the input data for each analysis frame by a known method to extract a predetermined S-order α parameter, and further uses this to derive an LSP coefficient as an S-dimensional space vector. To do. In this embodiment, S is an even number. The LSP coefficient is shown by the following equation (1).

= (P ₁ , P ₂ , ... P _s ) (1) P _{1 to} P _s shown in the equation (1) are expressed in the form of a set of resonance frequencies obtained based on the open / closed state of the sound path. It is a frequency domain parameter.

The LSP coefficient for each analysis frame obtained by the LSP analyzer 13 is supplied to the variable length frame selector 15.

The variable-length frame selector 15 performs variable-length frame selection based on piecewise optimal function approximation for the sequence of spectral envelope parameters input in analysis frame units as follows.

That is, in the present embodiment, 20 consecutive analysis frame sections, that is, 200 mSEC are set as one section, and the piecewise optimum function approximation is performed by the rectangle approximation for each section. In this case, it is of course possible to use another approximation function such as a polygonal line approximation function instead of the rectangular function as the approximation function.

Now, of the 20 analysis frames included in each category, i
It is assumed that the individual is selected as the representative analysis frame. Here, i is a number arbitrarily selected from 1 to 20. The LSP coefficient vector belonging to this representative analysis frame is also used to approximately represent LSP coefficient vectors belonging to other analysis frames in this section.In this case, a rectangular function approximation is performed and distortion due to this approximation error is performed. I is selected so that is minimized. The distortion in this case is basically obtained by using the vector distance of the above-mentioned coefficient as a scale, and the DP method is used based on the viewpoint that the sum of the vector distances between the coefficients of each analysis frame is the minimum. Can be easily requested.

In this way, the next i representative analysis frames are selected for each section, and the analysis frames to be represented by these i representative analysis frames are also determined and supplied to the multiplexer 17 as LSP coefficient data having a variable length frame structure. To be done.

The sound source information analyzer 14 analyzes the sound source information data by using a quantized speech signal for each analysis frame received from the window processor 12 to obtain a pitch period, voiced / unvoiced sound and sound source amplitude by a known analysis method. It is supplied to the multiplexer 17 for each frame.

The clock generator 16 provides the clock t ₁ required for digital processing from the A / D converter 11 to the multiplexer 17.

The multiplexer 17 sends the LSP coefficient data and the sound source information data input in this way to the decoding side 2 via the transmission path 100 as a multiplexed signal which is synthetically coded in a form suitable for transmission, and therefore also the information on the clock t _1. It is included in this multiplexed signal and provided to the decoding side 2.

On the decoding side 2, the demultiplexer 21 receives the multiplexed signal provided from the multiplexer 17 on the encoding side 1, extracts the clock t ₁ based on the spectrum of this time series data, and multiplexes while using this clock t _1. After the encoded signal is separated, it is decoded and then the multiplexer 17 on the encoding side 1
The LSP frequency data provided in the unit of division and the sound source information data provided in the unit of analysis frame are supplied to the frame synchronizer 22. The clock t ₁ extracted by the demultiplexer 21 is supplied to the frame synchronizer 22, the fixed length framing device 23, the memory 24, the correction message generator 26, etc., respectively.

The frame synchronizer 22 provides the clock t ₁ thus provided.
The information is received and supplied to the fixed-length framing device 23 while performing frame synchronization between the frame of the division unit in which the LSP coefficient data is transmitted and the analysis frame in which the sound source data is transmitted by using the information on the data.

The frame synchronizer 23 also extracts data relating to repeat bits from the LSP coefficient data received while performing frame synchronization, and power bits from the sound source data, and supplies these to the adaptive controller 25. The above-mentioned repeat bits are data relating to the number of analysis frames represented by each of the representative analysis frames in the i-th row selected for each section and the order within the section.

The fixed length framing device 23 uses the representative analysis frame and the repeat bit under the drive of the clock t ₁ provided from the demultiplexer 21, performs interpolation between the representative analysis frames, and fixes the LSP coefficient for each fixed frame. Restore the specified state. That is, the output state of the LSP analyzer 13 is restored.

In this way, the fixed length framed LSP coefficient is separately stored in the analysis frame unit in the memory 24 as a voice parameter together with the sound source information provided from the frame synchronizer 22 in the analysis frame unit.

The write clock when stored in the memory 24 is supplied from the demultiplexer 21 and the clock t ₁ indicated by the bold line is used. The clock read from this clock and supplied to the voice synthesizer 28 is supplied from the clock generator 27 on the decoding side 2. The supplied clock t ₂ shown by the dotted line is used. These two clocks are used for maintaining synchronization between the encoding side and the decoding side in the form of independent synchronization with each other. Such independent synchronization is excellent in terms of ensuring clock stability on the encoding side, a flexible system configuration, and efficient failure handling, but there is a call delay caused by a time lag between the two clocks. Inevitable. This will be described in detail below.

Now, assuming that the clock t ₂ on the decoding side 2 is deviated to a method faster than the clock t ₁ on the encoding side, the reading speed is faster than the writing to the memory 24 and the frame data to be finally read out. Will be lost. Further, when the clock t ₂ is shifted in the direction slower than the clock t _1, a state occurs in which a write command for the next frame data is issued before the reading is completed. In such a case, while monitoring the state of this deviation, when the deviation exceeds a certain allowable value, it is interleaved and shifted for writing and reading of the next analysis frame to ensure frame synchronization between the coding and decoding sides.

However, in the variable length frame type code decoding device,
The frame synchronization adjusted in this way processes one section in units of basic frames, and the call delay due to this is significantly increased as compared with a normal code decoding apparatus.

In the present embodiment, the above-mentioned frame synchronization correction is carried out in units of analysis frames which are sufficiently shorter than one segment, and the call delay is greatly improved.

Each time the fixed-length framing device 23 receives data regarding an adjustment frame from the adaptive controller 25, the fixed-length framing device 23 deletes the analysis frame from the analysis frames to be output to the memory 24. This adjustment frame has an essentially indefinite time length except for the consonants or the coexistence part of consonants and vowels, which has a strong influence on the reproduced sound quality during division, and therefore has a sufficiently small influence on the reproduced sound quality. It is assumed that the part and the silent part or the pause part in the voice which is effectively equivalent to the silent part are designated, and the timing thereof is set every time the modification message is received from the modification message generator 26.

By the way, in the spectrum envelope parameter of variable length frame, the time length of the analysis frame group designated by the representative analysis frame is long, and therefore the influence of one analysis frame deleted for maintaining frame synchronization is relatively small. One of the completed sections is the voiced sound stationary section, and the other is the silent section including pauses. When the voiced sound stationary part and the silent part are evaluated from the viewpoint of voice power, the former is obviously high in power and the latter is low in power. Which analysis frame is to be deleted is selected from the analysis frames designated by the repeat bits by using the power bit provided from the frame synchronizer 22 as follows.

The correction message generator 26 constantly receives the clocks t ₁ and t ₁ , and monitors the time lag ± Δt between the clocks. When this time difference ± Δt exceeds a certain allowable value, the correction message generator 26 outputs a correction message for causing the adaptive controller 25 to output the adjustment frame data designating the analysis frame to be deleted from the section.

Upon receiving the modification message, the adaptive controller 25 determines the temporally nearest deletion target analysis frame based on the power bit and the repeat bit, outputs this information as adjustment frame data, and holds synchronization in analysis frame units. .

The voice synthesizer 28 is supplied with the spectrum envelope parameter and the sound source data from the memory 23 in units of analysis frames while receiving the clock t ₂ .

The voice synthesizer 28 uses the LSP coefficient as a filter coefficient of the LSP synthesis filter, drives the filter with a sound source signal reproduced from the sound source data to generate a digital voice signal, and further applies this to a D / A converter to generate an analog voice signal. In addition to the signal, the LPF removes unnecessary high frequency components above 3.4 KHz to obtain the output voice.

In this way, variable length frame encoding and decoding can be performed in units of divisions, and frame synchronization retention can be processed in units of analysis frames.

In this embodiment, the time length of the section designated by the representative analysis frame among the spectrum envelope parameters and the time length of the silent section are used for adaptive control, but each of them is adaptively controlled independently. However, it does not matter.

Further, in the present embodiment, the analysis frame is used as a basic frame in synchronization as a section that is sufficiently shorter than one section, but it is clear that this can be easily performed even when the number of samples with almost the same time length is used. is there.

〔The invention's effect〕

As described above, according to the present invention, in the synchronization method of the variable-length frame type encoder that uses independent synchronization or forced synchronization for synchronization between the encoding side and the decoding side, the representative time length of the spectrum envelope parameter or silence By providing the means to adaptively control the section length or these two time lengths according to the degree of synchronization deviation between the encoding side and the decoding side, the call delay associated with the synchronization deviation correction is significantly reduced. There is an effect that it is possible to realize a synchronization method for a variable length frame type code decoding device.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention. 1 ... Encoding side, 2 ... Decoding side, 11 ... A / D converter, 12 ... Window processor, 13 ... LSP analyzer, 14 ... Sound source information analyzer, 15 ... Variable length frame Selector, 16 ... Clock generator, 17 ... Multiplexer, 21 ... Demultiplexer, 22 ... Frame synchronizer, 23 ... Fixed length framing device, 24 ... Memory, 25 ... Adaptive controller, 26 ...... Modified message generator, 26,27 …… Clock generator, 28 …… Speech synthesizer, 100 …… Transmission line.

Claims (1)

[Claims] 1. A synchronization method of a variable-length frame type code decoding apparatus which utilizes independent synchronization or forced synchronization for synchronization between an encoding side and a decoding side, and is designated by a representative spectrum envelope parameter for each variable-length frame section. If the decoding side clock precedes the encoding side clock, the voice analysis section or the silence section for each variable length frame section, or the time length of the voiced section and the silence section is extended by the basic analysis frame period to obtain the encoding side clock. Is preceded by the decoding side clock, it is provided with a synchronization holding means for performing synchronization control between the encoding side and the decoding side while adaptively controlling by shortening the basic analysis frame and correcting the synchronization shift. A method for synchronizing a variable length frame type code decoding device, comprising: