Investigation And Programmed On G.726 Speech Coding Algorithm

Telecommunication technology changes with each passing day, and it continually put forward new demands on speech coding. During the past 20 years, several standards of speech coding have been specified. Such achievements surpass the work done two decades ago. Currently, research focuses of industry include speech coding with low bit rate, low delay, low complexity and good perceptual quality.In this paper, we investigate speech waveform coding technology with emphasis on the G.726 Recommendation of ITU-T. Based on G.726, we present a new algorithm. Compared with G.726, our proposed algorithm achieves the same perceptual quality with lower bit rate. It has been implemented with VC++ on a PC platform.We first introduce the basic methods of speech processing in time domain. Emphatically, we describe linear prediction and tonality detection of speech signal. Moreover, we discuss the G.726 speech waveform coding standard in details. Then, we describe our improvement based on G.726.The G. 726 speech coding algorithm involves adaptive linear prediction to eliminate the correlation of signal waveform, so as to reduce the resulting bit rate. The adaptation processing includes linear prediction coefficient adaptation and adaptation of quantization step size for residual signals. Based on G. 726, we adopt a Huffman coder to make use of probability statistic of bit cascade covering N(N≥1) samples generated from ADPCM, in order to further reduce the bit rate.ng is lossless entropy coding, the speech quality of our improved algorithm should be same as that of G.726 standard.Experiment results also show that our improved algorithm achieves the same perceptual quality as G.726 standard while our improved algorithm uses a lower bit rate. Therefore, our proposal may lead to telecommunication bandwidth saving and storage requirement reduction.