Improving G.729 By Multi-stage VAD Algorithm

Speech communication plays an important role in modern communication world. VoIP is used all over the world with the development of Internet. However, the lack of band resource made low bit rate speech compression coding becoming a key part of speech communication. And one of the main goals of nowadays speech compression coding research is how to achieve best speech quality by least bit rate.In 1996 ITU-T issued G729, Coding of Speech at 8kb/s Using Conjugate-Structure Algebraic-Code-Excited Linear-Prediction (CS-ACELP), which is adopted by VoIP. G729 is the research object of this thesis because of its efficiency of compression and good quality of synthesis speech. But G729 is the most complexity algorithm that ITU-T ever proposed, its bit rate is fixed and it haven't compressed the silence time between effective speeches. Although G729B used VAD algorithm to compress the silence time, its VAD is too complicated and can't work efficiently when the SNR is lower than 10dB.This thesis adopts a two-stage VAD algorithm which won't add extra computing complexity to improve G729 algorithm and enhance its compression ratio. The two-stage VAD algorithm is composed of Short Time Average Magnitude Difference Function and Short Time Zero Crossing Rate Function or Short Time Autocorrelation Function. When the zero crossing rate of mix-speech is close to zero crossing rate of back noise, we use the Autocorrelation Function method as the detection method of the second stage. Or else, we use the Short Time Zero Crossing Rate as the second...