改进的基于小波包变换的语音特征提取算法

Improved algorithm of feature extraction based on wavelet packet for voice

针对语音信号的非平稳特性,传统的应用短时分析技术容易丢失信息的现状,提出了一种利用小波包变换的技术对语音信号的共振峰特征(FDWPT)进行提取的方法。对整个语音信号进行多分辨分析的小波包变换,这样可以得到每个频带的小波分解值,结合共振峰的频率特性,选取适当的小波包分解结点,对这些结点建立共振峰参数,使用矢量量化模型进行识别,从而提高了说话人识别的效果。

To solve the problem of causing information loss under the short-time analysis in the non-stationary characteristics,a kind of method to distinguish the Formant using Wavelet Packet Transform（FDWPT） is proposed.Firstly,the method applies wavelet packet transform to the speech signal.So a wavelet decomposing value is gotten and then combined with the formant features.The appropriate node is selected to establish the characteristic parameters of formant.Lastly VQ model is used to identify the parameters,and the effect of distinguish is enhanced greatly.