RussianPatents.com

Method and system for abolishing quantizer saturation during communication with data transfer in speech signal band. RU patent 2249860.

IPC classes for russian patent Method and system for abolishing quantizer saturation during communication with data transfer in speech signal band. RU patent 2249860. (RU 2249860):

G10L19 - Speech or audio signal analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, e.g. for compression or expansion, source-filter models or; psychoacoustic analysis

Another patents in same IPC classes:

Method for compaction and decompaction of speech messages / 2244963
Method comprises steps of preliminarily, at reception and transmission forming R matrices of allowed vectors, each matrix has dimension m2 x m1 of unit and zero elements; then from unidimensional analog speech signal forming initial matrix of N x N elements; converting received matrix to digital one; forming rectangular matrices with dimensions N x m and m x N being digital representation of initial matrix from elements of lines of permitted vectors; transmitting elements of those rectangular matrices through digital communication circuit; correcting errors at transmission side on base of testing matching of element groups of received rectangular matrices to line elements of preliminarily formed matrices of permitted vectors; then performing inverse operations for decompacting speech messages. Method is especially suitable for telephone calls by means of digital communication systems at rate 6 - 16 k bit/s.

Method and system for abolishing quantizer saturation during communication with data transfer in speech signal band / 2249860
Method and system for decreasing prediction error an averaging device for calculation of transfer coefficient is used, pulse detector, signals classifier, decision-taking means and transfer coefficient compensation device, wherein determining of compensated transfer coefficient of quantizer count is performed in process of coding/decoding of transferred data in speech signal band by use of vector linear non-adaptive predicting-type algorithm.

Improved spectrum transformation and convolution in sub-ranges spectrum / 2251795
Method for generating of high-frequency restored version of input signal of low-frequency range via high-frequency spectral restoration with use of digital system of filter banks is based on separation of input signal of low-frequency range via bank of filters for analysis to produce complex signals of sub-ranges in channels, receiving a row of serial complex signals of sub-ranges in channels of restoration range and correction of enveloping line for producing previously determined spectral enveloping line in restoration range, combining said row of signals via synthesis filter bank.

Method and device for reproducing speech signals and method for transferring said signals / 2255380
During encoding speech signals are separated on frames and separated signals are encoded on frame basis for output of encoding parameters like parameters of linear spectral couple, tone height, vocalized/non-vocalized signals or spectral amplitude. During calculation of altered parameters of encoding, encoding parameters are interpolated for calculation of altered encoding parameters, connected to temporal periods based on frames. During decoding harmonic waves and noise are synthesized on basis of altered encoding parameters and synthesized speech signals are selected.

Multi-mode encoding device / 2262748
Speech compression system provides encoding of speech signal into bits flow for later decoding for generation of synthesized speech, which contains full speed codec, half speed codec, one quarter speed codec and one eighth speed codec, which are selectively activated on basis of speed selection. Also, codecs of full and half speed are selectively activated on basis of type classification. Each codec is activated selectively for encoding and decoding speech signal for various speeds of transfer in bits, to accent different aspects of speech signal to increase total quality of synthesized speech signal.

Method for simulating auditory patient perception of acoustic signal after cochlear implantation / 2277375
Method involves applying analog-to-digital input signal transformation expressed as word, dividing transformed signal spectrum into odd and even frequency bands, summing odd bands, carrying out digital-to-analog transformation of resulting summed signal and training its perception by preliminarily getting familiar with the word shown for listening and following testing. Spectrum division is based on tonotopic frequency distribution law over cochlea axis. Frequency bands having odd numbers are arranged in equal distances along basilar membrane length in agreement with normal tonotopic frequency distribution law over cochlea axis. At least three odd spectrum bands are summed up. Training is carried out by multiple repetition of the word shown for listening until unambiguous correlation to the known word meaning given in preliminary acquaintance takes place. The same words are to be shown in testing and training.

Method and device for encoding an audio signal with usage of harmonics extraction / 2289858
In accordance to audio signal encoding method, harmonic components are extracted with usage of information resulting from fast Fourier transformation, which is received with usage of psycho-acoustic model 2 to received audio data of impulse-code modulation. Then, extracted harmonic components are removed from received audio data of impulse-code modulation. After that audio data, from which extracted harmonic components have been removed, are subjected to modified discontinuous cosine transformation and quantization.

Method and device for transmission of speech activity in distribution system of voice recognition / 2291499
Distributed system of voice recognition has voice recognition (VR) local mechanism in user unit and VR server mechanism in server. VR local mechanism has module for selection of features (FS), which selects features from voice signals. Voice activity detector (VAD) module detects voice activity invoice signal. Indication of voice activity is transmitted before features from user unit to server.

Method for analysis and synthesis of speech / 2296377
Method includes: analog-digital conversion of speech signal; segmentation of transformed signal onto elementary speech fragments; determining of vocalization of each fragment; determining, for each vocalized elementary speech segment, of main tone frequency and spectrum parameters; analysis and changing of spectrum parameters; and synthesis of speech sequence. Technical result is achieved because before synthesis, in vocalized segments periods of main tone of each such segment are adapted to zero starting phase by means of transferring digitization start moment in each period of main tone beyond the point of intersection of contouring line with zero amplitude, distortions appearing at joining lines of main tone periods are smoothed out and, during transformation of additional count in the end of modified period of main tone, re-digitization of such period is performed while preserving its original length.

Method for reverse filtration, method for synthesizing filtration, device for reverse filtration, device for synthesizing filtration and tools containing such devices / 2297049
In accordance to invention, filtration of input signal is performed for generation of first filtered signal; first filtered signal is combined with aforementioned input signal for production of difference signal, while stage of filtering of input signal for producing first filtered signal contains: stage of production of at least one delayed, amplified and filtered signal, and production stage contains: storage of signal, related to aforementioned input signal in a buffer; extraction of delayed signal from buffer, filtration of signal for forming at least one second filtered signal, while filtration is stable and causative; amplification of at least one signal by amplification coefficient, while method also contains production of aforementioned first filtered signal, basing on at least one aforementioned delayed, amplified and filtered signal.

FIELD: communication systems.

SUBSTANCE: method and system for decreasing prediction error an averaging device for calculation of transfer coefficient is used, pulse detector, signals classifier, decision-taking means and transfer coefficient compensation device, wherein determining of compensated transfer coefficient of quantizer count is performed in process of coding/decoding of transferred data in speech signal band by use of vector linear non-adaptive predicting-type algorithm.

EFFECT: higher efficiency.

4 cl, 4 dwg

THE TECHNICAL FIELD

The present invention relates, in General, to communication systems and, in particular, to the transmission of compressed signals in communication systems.

BACKGROUND OF INVENTION

Over the past years have been established in a variety of ways economical use of the required frequency, the ways in which communication using the transmission signals with compression reaches speech quality, relevant phone calls or close to the quality of telephone calls. These methods typically include the use of coding algorithms to narrow the frequency band required for transmission speeds of 64 kilobits per second without compression. One such example is the algorithm of the linear pre-compression in the code book with low latency (LPSK-M3) (LD-CELP), which allows you to narrow down the desired band of frequencies up to 16 kilobits per second. Naturally, to use such encoding algorithms both ends of the transmission path must be capable of encoding and decoding the transmitted signal. One of the solutions for the implementation of this requirement is to use the same private equipment at both ends of the transmission path and along it. Another possible solution is compliance with international standards, which ensure the compatibility of different types of equipment located along the transmission path.

International standard for encoding algorithm LPSC-M3 was published in March 1995 in the form of Recommendations G.728 sector for standardization International telecommunications Union (ITU) (ITU-T). However, it was found that this recommendation has some drawbacks. Among these disadvantages was the processing of the signal being transmitted at a variable speed transmission of binary data (referred to below as “SPDC” ("VBR")). In particular, this problem was noticed when using the Recommendation G.728 for data transmission in the band of the speech signal.

As a promotion for standardization sector of the International telecommunications Union (ITU) (ITU-T) company “III Telecom limited (ECI Telecom Ltd.) March 17, 1997 proposed solution, disclosed in Appendix J to Recommendation G.728 ITU. This publication, entitled “Algorithm with variable baud rate, especially designed for use LPSC-M3 during the data transmission in the band of the speech signal according to Recommendation G.728 ITU hardware multichannel data (AND) (DCME)&γτ; &γτ; ("Variable Bit-Rate algorithm, mainly for the Voice-band data applications of LD-CELP ITU-T Rec. G.728 in DCME"), incorporated here by reference. This publication will be referred to in future as “Algorithm 40 kbit/s”.

This publication describes solutions for PSPD and, in particular, used for data transmitted in the band of the speech signal (hereinafter referred to as "GPRS"). The publication presents information for the implementation of the codec corresponding to the algorithm LPSC-M3, as well as a modified version of the Annex G to Recommendation G.728, “Technical requirements for transmission speeds of 16 kbit/s with a fixed decimal point” to activate the mode switch to the arithmetic device with a fixed decimal point.

Codec described in the “Algorithm 40 kbit/s, essentially, use a data rate of 40 kbps. The duration of the algorithmic delay is equal to 5 samples, a total of 0.625 milliseconds, and the mode switching may be carried out through each "cycle adaptation" (2.5 MS).

The proposed algorithm 40 kbit/s was designed mainly to solve the transmission of compressed DPRS for applications such as AND, and it was suggested to replace them adaptive differential pulse code modulation (ADIM) with a speed of 40 kbps (recommendation G.726 ITU) in systems AND, which includes the algorithm LPSC-M3. Among the features provided by this algorithm includes a soft transition to the algorithm LPSC-MOH and from him and the quality of speech corresponding to phone calls or close to the quality of telephone calls.

Recommendation G.728, essentially, involves the presence of a cycle of adaptation used in the algorithm 40 kbit/s mode of transmission of speech. Therefore, when you return to the voice mode will be used LPSC-M3 specified in Recommendation G.728, and not the algorithm 40 kbit/s

The main modification of the codec operating in accordance with the algorithm 40 kbit/s, is the implementation of the principle of quantization with lattice encoding (referred to hereinafter as “KRK” ("TCQ")), described in Proceedings of the IEEE on Communications, vol. 38, No. 1 (1990) (IEEE Transactions on Communications, Vol.38, No.1 (1990)), which is included here by reference. This principle KRK replaces mode GPRS the principle of "analysis through synthesis" to search the code book that is listed in Recommendation G.728 ITU.

However, in the proposed algorithm 40 kbit/s had not been given the solution to the problem of how to avoid reaching saturation state when a pulse occurs in the error of prediction, for example, if there is a significant jump in energy level of the prediction errors. This problem leads to a high level of noise at the output of the decoding device and is known to be the cause of the mismatch between transmitter and receiver ends of the transmission path.

In U.S. patent 4677423 partly recognized the existence of such problems associated with a different type of algorithm, namely algorithm ADIGM, and opened the way to solve this problem. The mechanism described in U.S. patent 4677423, is one of the ways of overcoming the problems associated with an abrupt change of energy of signals in terms of frequency band, by fixing and removing the fixing speed of adaptation. Adaptation speed record in cases where the adaptation rate is very small, while the mode of removal of fixation used when you require a high speed of adaptation.

Unfortunately, because this solution is not fast enough for systems with coding algorithms, in which a prediction device is not adaptive, for example based on the analysis of linear prediction (referred to hereinafter as PL (LP)), then you need a different solution. For some tasks described in U.S. patent 4677423 solution is inefficient when trying to avoid saturation in the systems, including linear prediction device, when a pulse occurs in the forecast error. Some of these problems are the following: the solution according to the patent '423 based on the fact that the processing of each sample should be made separately, whereas in the linear prediction units instead of single samples, the proposed solution according to the patent '423, use a vector, which includes several samples, and the solution described in patent '423, is not sufficiently fast in order to use it in systems with linear prediction devices. Another main difference is that handled in the patent '423 errors are logarithmic errors that are unlikely to carry out the saturation of the quantizer so as quickly as linear errors. So you need a different solution that can answer this question for systems including linear prediction device.. the essence of the INVENTION

Therefore, the aim of the present invention is to provide a method for determining the offset of the conversion rate of the quantizer in the encoder using vector linear non-adaptive prediction algorithm method, which eliminated the above described disadvantages of the solutions of the prior art.

Another objective of the present invention is to provide a device of digital communication and the creation of a system that removes the problems caused by the presence of pulses in the error of prediction.

Further objectives and features of the invention will become apparent from the following description and accompanying drawings.

In accordance with the present invention, a method for determining the offset of the conversion rate of the quantizer in the process of encoding/decoding transmission GPRS type by using vector linear non-adaptive algorithm prediction type.

Mentioned below, the term "GPRS" is used to denote a modulated digital signals for transmission in the band of the speech signal (up to 4 kHz), for example, signals of modem signals DTMN (DTMF) DTMF) or signals of any other type with the same narrow band of frequencies.

The method proposed in accordance with the present invention, in the preferred embodiment includes the following operations:

I. create a vector of digital sample in coded form;

II. compute the coefficients PL for predicting the specified vector of digital sample and obtain from him the error vector linear prediction;

III. calculate the transmission coefficient of the specified vector error linear prediction;

IV. calculate the conversion factor quantizer based on the specified gear ratio;

V. on the basis of previous digital samples to calculate the average value of the transmission factor corresponding to the specified vector of digital sample;

VI. calculate the difference between the specified gear ratio and the specified average value;

VII. determine whether compensation transfer coefficient for momentum in the forecast error of the specified vector of digital sample on the basis of:

(a) comparison of the specified difference with a preset first threshold value, and

(b) comparing the difference between the transmission coefficients associated with the predefined number of most recently created vectors of digital samples and their respective average values and with a preset second threshold value;

VIII. if while performing the operation (VII) determined that the required compensation factor, determine the compensating value required for a pulse in the forecast error of the specified vector of digital sample;

IX. to get compensated for the conversion factor quantizer conversion rate quantizer obtained by performing the operation (V), summarize with a compensating value of the transmission coefficient, defined by the operation (VIII).

An example of such a non-adaptive linear prediction algorithm is an algorithm an all pole modeling.

Determining whether the signal is a strong signal, is carried out by comparing the differences existing between the transmission ratios associated with a pre-defined number of previous vectors digital sample, and the average values related to the same with a preset second threshold value. If the difference does not exceed this predetermined second threshold value, the signal can be considered a strong signal.

According to a preferred variant of the invention, the described method further includes the operation of calculating the value of the predetermined function, and the function formed by the calculated coefficients PL, corresponding to the vector of digital sample. The value thus obtained pre-defined functions can be used to determine the required compensation factor. According to this variant implementation this can be done, for example, by establishing restrictive conditions, namely, that up until the calculated value does not exceed a predetermined value, the compensation factor is not produced. Another possible example is the use of the compensation coefficient transfer coefficient, which depends on the difference existing between a calculated value and those for the pre-determined value.

An example of such a predetermined function according to this variant implementation of the function is equal to:

?AFTER (A[i])

i=1

where A[i] are the coefficients PL.

For any expert in the art it is clear that in this way can be used also other mechanisms for making decisions regarding the compensation of the transmission coefficient, and the results of their actions included in the final decision as to actually carry out compensation.

According to another variant implementation of the present invention, the maximum threshold value is predefined, and the calculated value of the difference calculated in operation (V) of the above method, compared with this maximum threshold value. This alternative implementation, along with other helps to extend the first predefined time period during which compensation gain up until its value is below a maximum threshold. The period of compensation transfer coefficient may be extended, for example, up until either short-term peak emissions will not fall below this maximum threshold, or within a longer predetermined period of time.

According to another preferred variant implementation of the present invention, the vector error of the linear prediction is obtained by performing quantization lattice code vector of prediction errors and the selection of the preferred quantized vector error linear prediction from multiple calculated quantized vectors of the error of the linear prediction. In the most preferred embodiment, this selection is performed by selecting the error vector linear prediction, which has the minimum error of prediction.

According to another variant implementation of the present invention when determining the required compensation factor according to the operation (VIII) impose restrictive threshold to prevent the production of excess compensation gain.

In accordance with another feature of the present invention it is proposed station digital communication for work in the digital communication system, comprising:

an input interface for receiving signals, data in the band of the speech signal and control;

the processing means designed to calculate:

coefficients PL for predicting the specified vector of digital sample and obtain from him the error vector linear prediction;

gear ratio specified the error vector linear prediction;

the conversion factor quantizer based on the specified gear ratio;

the average value of the specified gear ratio corresponding to the specified vector of digital sample over the previous digital samples;

the difference between the specified gear ratio and the specified average value;

first means for determining for determining whether compensation transfer coefficient for momentum in the forecast error of the specified vector of digital sample on the basis of:

A. compare the specified difference with a preset first threshold value, and

B. comparison of differences between coefficients of transmission associated with a pre-defined number of most recent vectors digital sampling, except that the specified created a vector of digital sample, and their respective average values and with a preset second threshold value,

the second determination tool designed to determine the required compensating values of the transfer coefficient to compensate for the pulse in the forecast error of the specified vector of digital sample if the determination made by the first means for determining is affirmative;

means summing multiplier quantizer with a compensating value of the transmission coefficient, which is defined by the specified second means of determining; and

an output interface for transmitting a data signal in the band of the speech signal.

To a person skilled in the art will understand that the above device may include additional features, which are essentially known from the prior art, and thus they are covered by the scope of the present invention.

It should be understood that as used below, the term "telecommunication network" covers the different types of networks known in the prior art, for example, the network NVR (TDM time division) (D), synchronous and asynchronous transmission, network, internetwork Protocol (IP)network, an IP network with frame relay and any other appropriate communication network.

The term "stations" is used here to describe a set of at least one pair of devices, encoding/decoding, one of which, when necessary, is used to convert the received signals into a new encoded form, and the other is used as a corresponding decoding device that converts signals received in this new encoded form, in the form that they essentially had to the coding device. These two devices can be either included in a single device, or may be separated from one another.

According to another variant embodiment of the invention it is proposed a communication device operating in the digital communication system, designed to create temporary change of the transmission factor for quantization in the encoding/decoding of the transmitted signal GPRS type, which includes the following devices:

I. averaging unit calculating a transmission coefficient;

II. the pulse detector;

III. classification of signals;

IV. the means of decision making and

V. compensation device of the transmission coefficient.

According to another preferred variant implementation of averaging the computing device operates in such a way that makes the calculation of the average estimated values of the transmission coefficient using the gear ratio for the most recent values of the vector and the difference G diff (G DIFF ) between the specified value of the coefficient of transmission of the last vector and the specified average value of the compensation factor. In the most preferred embodiment, receive the difference G DIFF (G DIFF ) and produce its comparison with a preset first threshold value by the pulse detector, which operates in such a manner that after a predetermined period of time discovers surges transfer coefficient.

According to another preferred variant implementation of the present invention, the signal classifier is arranged in such a way that detects predetermined transmitted signals GPRS, and in a more preferred embodiment, the means of decision making is arranged so that it receives the output signals of the pulse detector and classifier signals and accordingly actuates the device compensation gain.

In another preferred embodiment, the compensation device of the transmission coefficient functions in such a way that increases the gear ratio within a predetermined period of time.

On the other hand, in the proposed invention, the digital communication system for connecting several main telecommunication channels through a transmission path, comprising:

the first transmission medium located on at least the first end of the communication network for transmission of digital signals;

at least one pair of communication stations of the specified type, and

the means of reception, located on at least the second end of the communication network.

A BRIEF DESCRIPTION OF THE DRAWINGS

1 shows a diagram of the coding device, which includes a method of processing signals GPRS according to the present invention.

Figure 2 shows the layout of a typical device implementing States to generate charts lattice code.

Figure 3 presents a sample diagram of the lattice code generation which was used by devices implementing the state shown in figure 2.

Figure 4 shows the diagram of a method of implementation of the temporary change of the transmission factor for quantization in accordance with the present invention.

DETAILED DESCRIPTION OF THE INVENTION

Figure 1 presents a partial schematic structure of the coding device 1 of the present invention.

In the device 3 summation injected signal Sn together with its predicted evaluation value S n. Their difference is passed through a preamplifier 5 and serves to block 10 search KRK and making decisions on the Viterbi algorithm. The information received by this unit after processing the difference together with the corresponding input signal received from the block 12, which represents a set of extended ultra codebook is passed through the device 15 conversion gain and served in the prediction unit 16. All operations required for the algorithm CLC (quantization with lattice encoding) (TCQ) is performed in the device shown on the drawing by the block 10. These operations may include, for example, control of selected paths on the tree lattice code and preset values during playback, the calculation and comparison matrices and the computation of solutions of the Viterbi algorithm. Decisions on the Viterbi algorithm is a well-known state of the art and it is carried out according to the following procedure. Each node of a given set of nodes includes several acceptable branches. At each step of this procedure shall select a limited number of these branches, and the branches are the ones that will lead to the smallest error. After repeating this procedure for several samples produce the selection of the trajectory connecting the branches, which will lead to the minimum total error. In this configuration, the unit 10 generates 5 indexes channel marked in figure 1 by the letter j, by comparing the best chosen path Y j for 5 of the original samples by the Viterbi algorithm.

Figure 2 and 3 shows a diagram of a typical device implementing States, which performs the generation of the graph lattice code, and the graph lattice code.

In paragraph 7.1 "Algorithm 40 kbps" for each node specified permitted trajectory to the previous nodes through the lattice structure. For example, permitted the previous node to the first node (s [0] ) are node 0 in branch 0 (b [0]) and node 2 in branch 1 (b [1]).

In paragraph 7.2 "Algorithm 40 kbps" for each node specified permitted path to the next node through the lattice structure. For example, permitted the subsequent nodes to the first node (s [0]) are node 0 in branch 0 (b [0]) and node 2 in branch 1 (b [1]).

In paragraph 7.3 "Algorithm 40 kbps" specified subset of the quantization {D0, D1, D2, D3}, corresponding to each path through the lattice. For example, the transition from s[0] to s[0] corresponds to the subset DO. The transition from s[0] to s[l] corresponds to a subset of D2, and transitions to s[2] s[3] is not permitted and is therefore marked with X.

In paragraph 7.4 "Algorithm 40 kbps" provides the index bits, which marked each transition and defined two branches emanating from each node. For example, the transition from s[0] to s[0] is mapped (standing in a line, connected) 0. The transition from s[0] to s[1] corresponds to 1 (it should be note that the used bit 5, and h in equal 10h), a transitions to s[2] s[3] is not permitted and is therefore marked with X.

As indicated previously, the block 12 is introduced into the code book, which is a scalar quantizer Lloyd-max (Lloyd-Max) with the advanced set. 64 output level is divided into four subsets, beginning with the point that has the lowest negative value, and continuing to the point with the largest positive value, and successive points assigned to denote {D0, D1, D2, D3,..., D0, D1, D2, D3}. The quantization levels given in paragraph 7.6 "Algorithm 40 kbps", and the limiting values of the step described in paragraph 7.5 "Algorithm 40 kbps". The levels that belong to the subset of D0, shown in the column denoted by s [0]. Levels D1 shown under s [1],.., and D3 shown under s [3].

When the adapter 14 gear ratio in the reverse direction performs signal processing DPRS, according to the present invention there are a few differences in its functioning, compared with the method of processing speech signals in accordance with the standard G 728 ITU (G 728 ITU-T). The main differences are that:

1) In mode GPRS calculation of the RMS value of the output codebook is performed in the sequence of levels of the output signal (quantized difference), defined by the trajectory of the chosen path. RMS value is calculated over a sequence of 8 samples. However, unlike that disclosed in Appendix G to standard G.728, where pre-calculated tables remember the logarithm of the RMS value, mode GPRS necessary to carry out the calculation of the logarithm of the RMS value. Equation (1) gives a logarithmic approximation. The coefficients d 0 , d 1 , d 2 , d 3 , d 4, see paragraph 8 "Algorithm 40 kbps", and a detailed description of the logarithmic computing device described in paragraph 4.12.

Equation (I): 2· log10 (x) = d 0 · (x-1)+d 1 · (x-1) 2 +d 2 · (x-1) 3 +d 3 (x-1) 4 +d 4 · (x-1) 5 ,

where 1≤ x<2.

For values of x that are different from those listed above, perform the procedure of regulation. This procedure is described in the unit J.16 publication "Algorithm 40 kbps".

The output code book regarding the form and gear ratio, namely blocks No. G.93 and No. G.94 tables of logarithm of the transmission coefficient (the last two term in equation G-14) is replaced by the logarithm of the RMS value.

2) In the circuit taking the logarithm of the transmission coefficient can be entered smoothing filter, which reduces the steady-state oscillations for signals with constant variance, such as the data transfer signal in the band of the speech signal. To suppress both the speech signal and data signal, the quantizer algorithm with dynamic synchronization (DPT) ("DLQ") carry out the generation of adaptation to variable speed. Can be used in the DPT algorithm similar to the one described in recommendation G.726 ITU (ITU-T Rec. G.726).

Input processing device, which uses the algorithm DPT, serves the logarithm of d(n) gear ratio with fixed bias. This input signal average by weighing the filter (paragraph 4.13 "algorithm 40 kbps", unit No. J.14) for receiving the fixed coefficient G L transmission.

When a 1 =0, the quantizer is able to fully commit, and if a 1 =1 - in condition with completely unfixed. Calculating a 1 is carried out by comparing the energy of the quantized difference DT(n) over a long period of time and for a short period of time (paragraph 4.10, the unit J.12 "Algorithm 40 kbps"). Comparison characterizes the constancy of the variance of the quantized differences.

Equation (2): G = G U · α 1 +G L · (1-α 1 )

3) the Pulses of the prediction error can cause saturation of the quantizer. To prevent such a situation, according to the method proposed in the present invention, perform the temporal variation of the transmission factor for quantization. Naturally, for implementing the method of the present invention a preferred method for performing the calculation of the average value is the assignment when calculating the larger the weighting factor with the most recent values of the transfer coefficient.

Figure 4 shows the diagram of a method of implementation of the temporary change of the transmission factor for quantization. In accordance with this method produces the following operations:

A. Calculate the average value of the transmission coefficient:

In the smoothing filter 40 calculates an average value of evaluation values G ENVIRONMENTS (G AVE ) gear ratio using the most recent value GSTATE [0] transfer coefficient vector. In a preferred embodiment, the calculated average value is a weighted average, and the new values are assigned a higher weighting factor than the values obtained previously. In equation 3 shows the calculation of this average value, which is not mandatory. Then calculate the difference between the GSTATE [0] and G ENVIRONMENTS (G AVE ), which is denoted as G DIFF (G DIFF ), and serve it in a block 42 of the pulse detector.

Equation (3): G MEDIA = G POST · G ENVIRONMENTS +(1-G POST )· GS [0]

(G AVE = G CONCT · G AVE +(1-G CONCT )· GSTATE [0])

B. The block 42 of the pulse detector:

The function of this block is essentially the detection of a change gear ratio after a predetermined period of time during which the pulses were not detected. To accomplish this, make a comparison of G DIFF (G DIFF ) fixed preset second threshold value. If within a period of time exceeding a pre-set period of time, the value of G DIFF (G DIFF ) is less than a predetermined second threshold value, then the signal is considered "sustainable" signal. The detection pulse error linear prediction occurs when the value of G DIFF (G DIFF ) exceeds a predetermined first threshold value, and although it was determined that the previous signal is "sustainable" signal. According to a preferred variant implementation of the present invention a predetermined first threshold value is a predetermined second threshold value.

C. classification of signals:

In some cases, when the transfer GPRS pulses errors occur with the highest probability. So after their detection can be set to the maximum compensation parameters gear ratio. In block 44 the classification of these signals transmitted signals detect, for example, by use of the coefficients PL, and the classification process is passed to the block 46 decision making.

, Block 46 decision:

Block 46 decision accepts the output signal from block 44 of the signal classifier, and a block 42 of the pulse detector. Based on these output signals, decide whether you want compensation and what the impact will be described in the following paragraph, the compensation parameters gear ratio when the actuation unit 48 compensation gain.

D. Block 48 compensation transfer coefficient:

The main task performed by the block 48 is to determine the required compensation factor and to give the opportunity to increase the rate of transmission within a predetermined first time period. According to another variant embodiment of the invention, this predefined first time period may be variable. According to this other variant implementation establish a predetermined third threshold value for the threshold value of the maximum transfer rate. When a predetermined third threshold value is reached, the compensation factor is used over a long period of time in which this period may be redefined as a predefined second time period. The use of this alternative implementation may extend the period of compensation of the transmission coefficient in the case when the relative change of momentum is very large.

To a person skilled in the art will understand that in reality can be done in various variations and modifications of the method described above, whereby ultimately reach solutions to the same task and which fall under the scope of the present invention. For example, to achieve the desired effect instead extend the period of payment can be made by changing the level of compensation of the transmission coefficient. Besides, when to limit compensation use the limiter to ensure the best mode of carrying out the required compensation factor can be used the value set in this limiter.. the following is a description of other depicted in figure 1 units: 14 (adapter gear ratio in the reverse direction), 16 (prediction unit) and 18 (adapter coefficient inverse prediction).

A prediction device 16 is a shortened version of the synthesizing filter G.728 (block No. G.22). The order of the polynomial, which includes the coefficients PL, equal to 10 branches instead of the usual 50 branches used in synthesizing filter. The forecasting exercise based on the trajectory of the chosen path (paragraph 4.4, unit No. J.7 "Algorithm 40 kbps") in the following way: at time n form a prediction of the current sample for each node (paragraph 4.5, the unit J.8 "Algorithm 40 kbps") using the sequence of play, which is set to the selected path at time n-1. When you use this method, perform only step-by-step scalar prediction and the prediction is not substantially extend into the future. This makes the prediction more "limited"than in many other predictive VQ schemes.

The adapter 18 of the coefficient of the inverse prediction is similar to the adapter filter reverse synthesis (block No. G.23). The main differences are the following:

- Perform the calculation only 10 parameters CLP (coding linear prediction (LPC)). Hybrid module processing method open (unit G.49) continuously performs the calculation of the 51-th autocorrelation coefficient, improving performance in the transitions from data to speech signal.

The coefficient of expansion of the band synthesizing filter now equal 240/256. Expansion coefficients of the frequency bands listed in paragraph 9 "Algorithm 40 kbps".

EXAMPLE:

To assess the efficiency of the method proposed in the present invention, was performed the following set of experiments. Was evaluated transmission GPRS Protocol (V.23 supported in character mode using the algorithm of 40 kbit/s G.728. When the assessment was made on the comparison of the transmitted symbols with the accepted and was carried out by counting the number of mismatches found in the total number of transmitted symbols. This attitude characterized the average error.

When you use the recommendation G.728, including amendment "Algorithm 40 kbps", it was found that the average error is equal to approximately 33%.

By similar experiments were evaluated method proposed in the present invention.

The values of the first and second predetermined threshold values have been pre-set to 1800. When he found that the pulse is generated when the predicted gear ratio exceeds 1800, it is provided that the previous 80 vectors of digital samples, each of which consisted of 5 samples, where each sample had a duration of 125 μs, were identified as signals "sustainable" type trigger mechanism of compensation prediction. Observed a striking reduction defined above, the average error and its reduction to about 0.05%.

It should be understood that the above description is only to illustrate some embodiments of the invention. Specialist in the art can think of many other methods of implementing the invention, without departing from the invention, and thus they fall under the scope of the present invention.

1. The communications device is operating in the digital communication system and configured to determine the offset of the conversion rate of the quantizer in the process of encoding/decoding data type transmitted in the band of the speech signal (DPS), through the use of vector linear non-adaptive algorithm prediction type containing

2. The device according to claim 1, characterized in that the signal GPRS is considered a stable signal when the difference between the coefficients of transmission associated with a pre-defined number of previous vectors digital sample, and the average values related to the same, with a preset second threshold value, does not exceed the specified predetermined second threshold value.

3. The device according to claim 1, wherein the processor is additionally configured to calculate the value of a predetermined function, which is made by the calculated coefficients PL associated with the specified vector of digital sample.

4. The device according to claim 3, characterized in that the predefined function is used when determining the required compensation factor.

5. The device according to claim 3, characterized in that the predefined function is equal to:

6. The device according to claim 1, characterized in that the error vector of linear prediction is obtained by performing quantization lattice code vector of prediction errors and the selection of the preferred quantized vector error linear prediction from multiple calculated quantized vectors of the error of the linear prediction.

7. The device according to claim 6, characterized in that the selection is carried out by selecting the error vector linear prediction, which has the minimum error of prediction.

8. The communications device is operating in the digital communication system and configured to create a temporary change of the transmission factor for quantization in the encoding/decoding of the transmitted signal type data transmitted in the band of the speech signal (DPRS)containing

9. The communication device of claim 8, wherein receiving the specified difference G diff and its comparison with a preset first threshold value is realized by means of the specified pulse detector, which operates in such a manner that after a predetermined period of time produces the detection of abrupt changes in gear ratio.

10. The digital communication station operating in the digital communication system containing

11. The digital communication system for connecting several main telecommunication channels through a transmission path containing