基于概率神经网络说话人识别的算法研究

Study of PNN speaker recognition

为实现由语音信号进行说话人身份的辨识,研究了以往的实现说话人辨认的系统,提出一种改进的算法,采用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,即基于概率神经网络(PNN)的识别方法。经实测数据的处理,表明PNN对训练样本有很高的分类准确率,且对测试样本的分类准确性也较高,并验证了本方法的有效性。

In order to make identification from the speech signal, based on analysis of the conventional identical algorithm, it proposes an advanced method, which uses Mel Frequency Ceptral Coefficients （MFCC） as feature parameters. That is the method of probabilistic neural Network （PNN）. The simulation shows that, the classification accuracy rate of PNN to the training samples is very high, and the classification accuracy rate of PNN to the testing samples is high either, the real data testified the effectiveness of this proposed algorithm.