Speech Signal Classification Based on Convolutional Neural Networks

In the field of intelligent human-computer interaction, speech signal is the hotspot research field, and has been widely used. For the traditional classification algorithm, the computational complexity is high and the classification accuracy is low. This paper proposes a convolutional neural network based on convolutional neural network. The speech signal classification method converts the speech signal into a form of a spectrogram and inputs it into a convolutional neural network to realize classification of the speech signal. Finally, the training and testing of convolutional neural networks are completed by using the framework of tensorflow. Compared with the traditional classification algorithm, the accuracy of the classification algorithm proposed in this paper reaches about 98%. The results show the feasibility and effectiveness of the experimental method.