A Simplified Method Of HELP Codec

With modern electronic technology, communication technology and communication are rapidly expanding the scope, the noise environment has been surprised by voice communication, and speech signal will certainly be greatly influenced by the surrounding environment. But now the quality of life people have more and more requirements, which the researchers made the speech more serious issue. The main voice processing technology of modern coding methods are: parameter coding, waveform coding, hybrid coding [1] [3] [10].In the parameter encoding, speech parameter extraction to quantify the accuracy and encoding parameters directly affect the quality of voice communication, but still not a standard and accurate parameter extraction method, so the extract from the noisy speech pitch period, Extraction of the linear prediction coefficients describe the channel, and effective encoding parameters have important research value and practical applications.Pitch is exceptionally important speech coding parameters of excitation source, usually before the extraction, should be the appropriate pre-processing, filtering through a filter low-frequency and high-frequency noise, clipping level again , highlighting the cyclical. From the use of perspective, is proposed based on autocorrelation and CEP-strum analysis combined with the rapid extraction method of pitch period, through the voice signal from the correlation operation, highlighting the peak amplitude with periodic, periodic reduced amplitude, enhance the accuracy of pitch extraction, but because of the resonance peak amplitude sometimes over a period range, which may be periodic phenomenon of frequency doubling and a half [2] [6], in order to solve this problem, and then the autocorrelation signal for CEP-strum computation to avoid such problems, so that the cycle of extraction more accurate.In this paper, the parameters of coding speech coding, LPC is given based on the harmonic analysis of the linear prediction coefficient (HELP codec) technology. Harmonic is the most intuitive voice signal model, it is more easy to understand, and linear predictive analysis is the ideal excitation source, which together can make the synthesized speech quality significantly improved.