RU2400832C2 - Method for generation of excitation signal in low-speed vocoders with linear prediction - Google Patents

Abstract

FIELD: information technologies.

SUBSTANCE: method is realised through rejection to send information on excitation signal along information channel, which is synthesised directly at the moment of reception by data on parameters of synthesising model. Information on parameters of synthesising filter comprising information on coefficients of prediction or linear spectral frequencies is sent along communication channel, as well as value of excitation signal amplification coefficient. At the moment of reception these parameters are used to calculate amplitude-frequency characteristic of synthesising filter at phase angles of its poles, and spectrum of amplitudes and phases of excitation signal is generated. Signal of excitation is formed on the basis of data on spectra of its amplitudes and phases and amplification ratio.

EFFECT: reduced speed of data transfer along communication channels due to efficient coding of speech signals on the basis of linear prediction.

2 dwg

Description

The invention relates to the field of digital communications and can be used in telecommunication communication systems for efficient coding of speech signals.

The main problem of the digital representation of the speech signal is the task of high-quality and compact data coding for their transmission via digital communication channels. The solution to this problem will allow, under the conditions of a specified communication quality criterion, to increase the throughput of linear paths and transmission channels. Often, in some tasks of encoding a speech signal, it is supposed to reduce the transmission rate while maintaining qualitative indicators of its perception.

Among the variety of methods for encoding speech signals, one of the most effective is the linear prediction method. The linear speech prediction method belongs to the class of methods using the speech signal model in the form of a response of a linear system with variable parameters (voice path) to the corresponding excitation signal (generating signal). The analyzer of the speech-converting device extracts from the short segment of the speech signal the state parameters of the linear system and the excitation signal, allowing the synthesizer to restore the original signal with the required degree of fidelity.

Known methods for the synthesis of excitation signals in vocoders with linear prediction, based on the analysis of the remainder of linear prediction (O.I. Shelukhin, NF Lukyantsev. Digital processing and transmission of speech. M., Radio and Communication, 2000 - S. 102 -166; Rabiner L.R., Schafer R.V. Digital processing of speech signals. - M .: Radio and communications, 1981. - S.365-428).

In devices that implement these methods, the analysis of the remainder of the linear prediction is performed to synthesize the excitation signals at the reception. In this case, information on the coefficients of the forming model, the parameters characterizing the excitation signal and the encoded speech signal is transmitted through the communication channel.

To effectively present the excitation signal in such devices, various methods are used to reduce the informative redundancy of these sequences (US patents No. 7289952 dated 10/30/2007, US No. 7233896 dated July 10, 2007, US No. 7133823 dated November 7, 2006, US No. 5963897 dated October 5, 1999, US No. 6757650 dated June 29, 2004, RU No. 2133399 dated October 22, 1997, RU No. 97117357 dated February 20, 2001).

The disadvantage of these methods is the significant expenditure of information resource on the presentation of the excitation signal when it is transmitted over the communication channel, and, therefore, the relatively high speed of data transmission over communication channels with effective speech coding.

The technical result of the proposed method is to reduce the data transfer rate over communication channels with efficient coding of speech signals based on linear prediction.

To achieve this technical result in a linear prediction vocoder, it is proposed to exclude the transmission of information about the excitation signal over the communication channel. The formation of the excitation signal can be implemented directly at the reception according to the data on the synthesizing model. At the same time, information on the parameters of the forming model, the gain and the parameters characterizing the encoded speech signal, which should be calculated on each quasi-stationary segment of the analysis of the speech signal, is transmitted through the communication channel.

Experimental studies of the linear speech prediction model showed that if we analyze the remainder of the linear prediction, then a practically slightly distorted speech signal is clearly traced. This fact indicates the similarity of the normalized power spectral densities of the real speech signal and the remainder of the linear prediction, and further studies have confirmed this assumption. A recursive filter is used as a synthesizing filter in linear prediction, while its poles reflect the maxima of the generated amplitude-frequency characteristic (AFC), which correspond to the maxima of the spectral power density of the analyzed segment of the speech signal in the quasistationary section. The parameters of the synthesizing system in linear prediction reflect the formant structure of speech in the quasistationary section. As such parameters, linear spectral frequencies are used. In this case, it is sufficient to transmit and receive data on the values of linear spectral frequencies and their amplitudes. The excitation signal in such a linear prediction vocoder is formed as a superposition of formant frequencies of the spectral power density of the speech signal in the quasistationary section. In this case, the initial phases of the corresponding frequencies are calculated under the assumption of a uniform distribution. In the literature devoted to the analysis and perception of a speech signal, it is pointed out that the information about the phase spectrum of a speech signal is not essential for its intelligibility and does not significantly affect its perception (Markel J.D., A.Kh. Gray. Linear speech prediction . - M.: Communication, 1980. - S.166-196).

The essence of the proposed method is as follows. To generate the excitation signal in the vocoder based on linear prediction at the reception, the parameters of the synthesizing filter containing information about the prediction coefficients or linear spectral frequencies, as well as the value of the gain of the excitation signal, are extracted from the transmission frame. According to these parameters, the frequency response of the synthesizing filter is calculated at the phase angles of its poles and a spectrum of amplitudes and phases of the excitation signal is formed. Then, an excitation signal is generated based on data on the gain and spectra of its amplitudes and phases, which is used in the synthesis filter of the lipreader to form a digital speech signal in the quasistationary region.

The functioning algorithm of the proposed system that implements the method of generating an excitation signal in low-speed vocoders with linear prediction is presented in figure 1.

The analysis of the prior art made it possible to establish that analogues that are characterized by a combination of features that are identical to all the features of the claimed technical solution are absent, which indicates the compliance of the claimed method with the condition of patentability "novelty".

Thanks to a new set of essential features of the system, which ensure the exclusion of information about the excitation signal from the transmission frame format and its formation at the reception by the values of the parameters of the synthesizing model, a significant decrease in the data transfer rate in the communication channel is achieved.

Analysis of existing technical solutions in this area showed that the introduced distinguishing features are absent and do not follow explicitly from the prior art. Therefore, the claimed technical solution meets the criterion of "inventive step".

The industrial applicability of the introduced elements is due to the presence of the element base on the basis of which they can be performed.

The claimed technical solution is illustrated by the drawing (figure 2), which shows a functional diagram of a device that implements a method of generating an excitation signal in low-speed vocoders with linear prediction by using information about the parameters of the synthesizing model.

A device that implements this method consists of a block for receiving a transmission frame of a vocoder with linear prediction 1, which directly receives the information sequence of the transmission frame. It is connected to the unit for selecting the parameters of the synthesizing filter 2, in which the set of parameters describing this filter is formed. The generated parameters are fed to the block for calculating the frequency response of the synthesizing filter at the phase angles of its poles 3. This block is connected to the block for generating the excitation signal 4. The signal from the block for generating the initial phases of the spectral components of the excitation signal 5 is supplied to the other inputs of the block unit 1, which contains additional information for generating an excitation signal, for example, a gain value. The generated excitation signal is fed to block 6, the receiver of the excitation signal of the low-speed linear prediction vocoder, where it is used to excite the synthesis filter of the receiving part of the linear prediction vocoder.

A device that implements the claimed method works as follows. A linear prediction vocoder transmission frame formed on the quasi-stationary part of the speech signal is sent to block 1, in which it is analyzed and the information component describing the forming filter is selected. This signal is sent to block 2, in which the values of the poles of the transfer function of the synthesizing filter are calculated, according to which in block 3 the values of its frequency response at the corresponding phase angles are calculated. In block 4, the excitation signal of the synthesizing filter is formed according to its amplitude spectrum obtained from block 3, the phase spectrum, and gain obtained from blocks 5 and 1, respectively. This excitation signal is then fed to block 6, where it is used to excite the synthesizing filter of the receiving part of the vocoder with linear prediction.

The advantages of the method include the fact that the elimination of information on the excitation signal from the transmission frame can significantly reduce the data rate in the communication channel, as well as reduce the computational complexity of the speech signal encoding algorithm on the transmitting side. Using the proposed technical solution for generating an excitation signal in vocoders with linear prediction will reduce the data transfer rate in the communication channel by 40-50% of the known solutions, or redistribute the information resource provided by the communication channel to the formation of additional subscription services.

The above technical solutions show that a device embodying the invention in its implementation is capable of providing a lower data transfer rate by eliminating information about the excitation signal from the transmission frame and generating it at the reception according to the parameters describing the voice path model.

Claims (1)

A method for synthesizing excitation signals in linear prediction vocoders based on generating an excitation signal by using parameters of a synthesizing filter, characterized in that for synthesizing an excitation signal in vocoder based on linear prediction at the reception, synthesizing filter parameters containing information on prediction coefficients are extracted from a transmission frame or linear spectral frequencies, as well as the value of the gain of the excitation signal, according to these parameters, am the frequency-response curve of the synthesizing filter at the phase angles of its poles and form the spectrum of amplitudes and phases of the excitation signal, then generate the excitation signal based on the gain and spectra of its amplitudes and phases, which is used in the synthesizing filter of the vocoder with linear prediction to form a speech segment signal in the quasistationary section.

Images