基于MFCC的说话人语音识别系统的研究

摘要

说话人识别是当前语音识别的研究热点之一。本文主要研究了以下几个方面：说话人语音识别系统，对能够反映人对语音感知特性的Mel频率倒谱系数(MFCC)作为特征参数进行提取。同时，分析了概率神经网络PNN，概率神经网络是性能良好的分类神经网络。实验结果表明，概率神经网络PNN对训练的语音样本有着很高的分类准确率。%Speaker recognition is one of the hot topics in speech recognition. This paper mainly studies the following aspects: Speak-er recognition system, which can reflect the characteristics of the speech perception of the Mel frequency coefficients (MFCC) as a fea-ture parameter extraction. At the same time, it analyzes the probabilistic neural network PNN, and the performance of the neural network is good. The experimental results show that the PNN can have very high classification accuracy for the training of speech samples.