基于动静态组合特征参数的语音识别  

Speech Recognition Based on Dynamic and Static Feature Parameter Integration

作　　者：王旭[1] 韩志艳[1] 王健[1] 薛丽芳[1] 

机构地区：[1]东北大学信息科学与工程学院,沈阳110004

出　　处：《计算机科学》2008年第7期129-132,共4页Computer Science

基　　金：国家自然科学基金重大资助项目(50477015)

摘　　要：基于语音信号的时变特性,本文提出了动静态特征参数结合的语音信号识别方法,首先在特征参数提取中引入了小波包变换,借助MFCC(Mel-Frequency Cepstrum Coefficient)参数的提取方法,用小波包变换代替傅立叶变换和Mel滤波器组,提取了新的静态特征参数DWPTMFCC(Discrete Wavelet Packet Transform Mel-Frequency Coeffi-cient),然后把它与一阶DWPTMFCC差分参数相结合成一个向量,作为一帧语音信号的参数,通过试验和仿真,此参数具有很高的识别率,是一种很好的语音特征参数。并且把混沌特性引入到神经元,构成混沌神经网络,把这种神经网络用于语音识别,并与常用的BP神经网络识别方法进行了比较。试验结果表明,混沌神经网络的平均识别率要高于同等条件下常用的神经网络方法的识别率。In this paper,we propose a new speech recognition method of dynamic and static feature integration. Wavelet packet transformation method is introduced to feature parameters in virtue of MFCC（Mel-Frequency Cepstrum Coefficient）, then combined difference feature formed dynamic and static feature parameter integration. The simulation results indicate that the new feature parameter is better than MFCC, that is a very nice speech feature parameter. At the same time, chaotic dynamics is introduced to neurons, and a chaotic neural network（CNN） is built. And compared the performance of the network with BP neural network, the experimental results show that chaotic neural network method outperforms the BP method.

关 键 词：语音识别 特征参数 小波包变换 混沌神经网络 

分 类 号：TP223[自动化与计算机技术—检测技术与自动化装置] TN912.34[自动化与计算机技术—控制科学与工程]