基于卷积神经网络的语音特征提取算法

Speech recognition as a way of human-computer interaction, so prevalent in today's computer age, how to make intelligent computer can accept human oral instruction, understand the human language is a very challenging and meaningful topic. Speech recognition and other machine learning problem, feature selection is also one of the most important link of starting, one has to distinguish between sex and stability is characterized by a classification problem is the premise of good recognition rate. With the increase of Internet information grows exponentially, huge amounts of voice and data features has a great deal of speaker-independent, difference and noise interference methods of feature extraction and feature transformation is hard to meet the needs of the current training model identification. In recent years based on speech recognition and deep learning theory together, through the study found that the convolutional neural network structure is very suitable for speech signal feature extraction process, this paper proposes a feature extraction method based on convolution neural network, and the combination of relatively complex GMM - HMM model of the new voice recognition system. Experiments show that the convolution neural network structure can be very good to overcome the differences between speaker-independent speech signals and the influence of noise, GMM - HMM model is more relevant than softmax classifier in speech complex signal model area, the final recognition rate had the very big improvement.