Research On Speech Endpoint Detection Algorithm And Implementation In 3G

Speech endpoint detection can reduce computation in real-time system largely. Therefore, the system can deal with speech input only .and avoid wasting computing and storing in silence segment. in favor of implementation in DSP system short of resources. This paper begins with an analysis of VAD(Voice Activity Detector) algorithm in AMR Vocoder(speech codec standard in WCDMA), with emphasis on the feature and fundamental principles of it, followed by a revised method that have better performance in simulation testo and is proved practicable in experimental test . Moreover, the paper discusses the principle of software and hardware design , based on the implementation of Multi-channel AIvIR Vocoder on the TMS320C6203. The results obtained from emulation are in bit exact agreement with the calculated results provided by 3GPP. The informal subjective test in real-time processing indicates that the synthetic speech quality of AMR Vocoder is better than it of RPE-LTP in GSM system, achieving toll quality, that can apply to devices employing the AMR Vocoder within the 3GPP system in terms of software and hardware.