【作者】 岳子琪；

【导师】 张海林；

【作者基本信息】 西安电子科技大学 ， 通信与信息系统， 2002， 硕士

【摘要】 语音端点检测可以减少实时系统中的大量计算，使该系统仅处理语音输入，不至于在静音段白白浪费计算量和存储量，有利于在资源较为紧张的DSP系统上实现。本文首先对WCDMA的语音编码标准—AMR声码器中的语音端点检测算法进行了分析，重点探讨了算法的特点及其理论依据，其思想具有普遍意义，可以在信源压缩编码及其它需要高精度语音端点检测的应用场合中采用，具有较大的研究价值。然后，本文进一步提出改进的语音端点检测算法，仿真数据表明性能有较大提高，且在项目中得以实际运用，测试性能良好。本项目实现了基于TMS320C6203的灵活高效多通道AMR声码器，仿真和录放验证的结果与3GPP提供的结果满足比特级精确要求。实时处理的非正式主观测试表明，合成语音质量优于GSM的RPE-LTP的语音质量，达到长途语音质量，完全可以实际应用，为第三代移动通信中声码器设备的研制奠定了良好的软件和硬件基础。更多还原

【Abstract】 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.更多还原