Abstract: The feature coefficients based on MFCC are not fully reflecting speech information as a result of speech signal movement and overlap of frames，especially noisy effect.A new method for noise robust speech recognition based on a hybrid model of Hidden Markov Models（HMM） and Wavelet Neural Network（WNN） is presented.The model trained by this method is used in MFCC coefficients.Experimental results show this model has better noise robustness.

摘要：

通过MFFC计算出的语音特征系数，由于语音信号的动态性，帧之间有重叠，噪声的影响，使特征系数不能完全反映出语音的信息。提出一种隐马尔可夫模型（HMM）和小波神经网络（WNN）混合模型的抗噪语音识别方法。该方法对MFCC特征系数利用小波神经网络进行训练，得到新的MFCC特征系数。实验结果表明，在噪声环境下，该混合模型比单纯HMM具有更强的噪声鲁棒性，明显改善了语音识别系统的性能。