与文本无关的声纹识别系统的研究

Research on Text-independence for Voiceprint Recognition System

声纹识别按识别的方式分为与文本相关和与文本无关两类。针对声纹识别技术中与文本无关的声纹识别问题进行研究。为提高系统的识别率,提出并实现了多特征参数组合的识别算法。该算法以LPCC,MFCC组合作为特征参数并将其应用到声纹识别系统中,实验结果表明组合特征参数由于充分利用了语音信号的相关特性和人耳听觉感知特性,比单独使用一种参数具有更好的识别效果。

Voiceprint recognition was divided into text-dependence and text-independence according to the recognition pattern. In view of voiceprint recognition technique, the problem on the text- independent voiceprint recognition was resear ched. Multi-feature vectors combination recognition algorithm was proposed and implemented to improve the system recognition rate. This algorithm took the combination of LPCC and MFCC as the feature parameters which were also put into use in voiceprint recognition system. The result of experiments expatiates that the parameters featured combination has much better recognition effect than a single kind of parameters used, because relative voice information with features and human hearing characteristics are fully used.