A feature extraction method based on Gauss wavelet filter and combined wavelets filter in speech recognition

This paper used optimized frame algorithm in noise-robust feature extraction of speech recognition and introduced a feature extraction method based on optimized frame algorithm. The method of calculating the sample numbers of observation window in ZCPA feature extraction was proposed by studying the length of frame and human auditory characteristic. This paper also discussed the effect of using different frame in ZCPA feature extraction in detail. The RBF neural net was used in training and recognition course. The results showed that new feature has higher recognition rate and better robustness than traditional feature.