基于高斯小波滤波器的语音识别特征提取方法

A Feature Extraction Method Based on Gauss Wavelet Filter in Speech Recognition

把小波理论应用于抗噪语音识别特征提取,提出了基于高斯小波滤波器的语音识别特征提取方法,通过对人耳听觉特性的研究,按照人耳临界带宽设计了一组高斯小波带通滤波器。详细讨论了高斯小波滤波器的尺度参数选择方法。使用RBF识别网络,仿真实现了使用新特征与原特征的识别结果,证明了新特征具有较高的识别率和优良的抗噪性能。

This paper uses wavelet theory in noise-robust feature extraction of speech recognition and introduces a feature extraction method based on Gauss wavelet filter. The Gauss wavelet filter with human critical frequency band is obtained by studying human auditory characteristics. This paper also studies the method of scale parameter choosing in designing Guass wavelet filter. The methods with new and original feature are simulated. The RBF neural net was used in train and recognition course. The results show that new feature has higher recognition rate and better robustness than traditional feature.