Research On Preprocessing Algorithm Of G.722 Wideband Speech Signal

With the development of economy globalization, people's life will not be restricted in one city or one country. More and more companies communicate with each other using video and audio conference. So wideband digital speech communication with high quality becomes more and more desired. Now in the area of wideband digital speech communication, G.722 standard which was approved by CCITT in 1988 becomes more and more popular. G.722 standard whose bandwidth is 7000Hz and bit rate is 64kbps, extends the transmission bands from 200～3200 Hz to 50～7000 Hz. Also the speech quality of G.722 is much better than traditional narrow band codec. This also makes it possible for transmitting high quality music signal. There are three kinds of bit rate for G.722 codec. They are 64/56/48 kbps. When the network is overload, G.722 codec will automatically decrease the bit rate to 56/48 kbps to save the bandwidth. Also when network becomes normal, bit rate of G.722 codec will recover to 64 kbps. This will make use of the bandwidth much more efficiently and make those applications, such as video and audio conference which base on G.722 codec, much more reliable. Based on these, this paper includes these aspects of research.Firstly, this thesis introduces history and actuality of wideband speech codec. Also it gives some basic concept about speech preprocessing.Secondly, it introduces some theory which is related with speech preprocessing, such as masking effect, effect of frequencies for human sensation and also homostasis filting theory in adaptive mel-cepstral analysis.Thirdly, after research on the principles of G.722 speech codec and functions of each module, the paper research and discuss several algorithms, such as quantized noise shaping algorithm, adaptive pastfiltering algorithm, adaptive mel cepstral analysis, long time predication and so on.Lastly, this paper gives an algorithm which applies quantized noise shaping algorithm and adaptive postfiltering algorithm on G.722 codec. Also it gives the simulation result to verify the performance, advantage and disadvantage of this algorithm according to these features, such as wave form in time domain, spectrum density and power spectrum density and so on.After software simulation and verification, the algorithm which is proposed in this paper will not increase the complexity of G.722 codec, and PESQ score will not decrease, especially the subjective quality of speech will improve in some aspects.