论文部分内容阅读
讨论了以单一矢量量化和基于矢量量化动态规划两种算法为基础的说话人证实方案，简明介绍了语音特征提取，叙述了矢量量化动态规划基本算法。计算机仿真结果表明，通过对每个人的个人性语音信息的分析，矢量量化动态规划算法不但能相当准确地证实一个人的身份，而且能识别出冒名顶替者；此算法在ＤＳＰ（数字信号处理）上所实现的实时说话人证实系统的证实错误率已达到实际应用水准。
Discusses the speaker verification scheme based on single vector quantization and vector quantization based dynamic programming. It briefly introduces the speech feature extraction and describes the basic algorithm of vector quantization dynamic programming. Computer simulation results show that, through the analysis of individual personal voice information, vector quantization dynamic programming not only can confirm a person’s identity quite accurately, but also can recognize the impostors; this algorithm in DSP (Digital Signal Processing) Realization of real-time speaker verification system confirmed the error rate has reached the practical application level.