Research And Implementation Of Speech Emotion Recognition Based On Convolutional Neural Networks

In recent years,with the rapid development of artificial intelligence technology,human-machine voice interaction has become more intelligent.Speech emotion recognition in human-computer interaction has been a hot topic in the field of artificial intelligence.It makes robots have the same emotional state as humans,which is an inevitable trend in the future.This paper mainly studies the speech emotion recognition algorithm based on convolutional neural networks.In terms of the convolutional neural network algorithm,this paper firstly introduces the basic theory of convolutional neural network,and then improves the convolution kernel weight updating algorithm in the process of the convolutional neural network training.So that the update of the convolution kernel weight value is related to the number of iterations.It improves the expression ability of convolutional neural networks.In terms of speech emotion feature extraction,this paper introduces commonly used speech feature extraction methods and selects the most widely used Mel-Frequency Cepstral Coefficients(MFCC)method for extracting speech features.In experiments,in order to increase the feature differences between emotional voices,the MFCC characteristic data matrix obtained by speech frequency signal after preprocessing is transformed to improve speech emotion recognition rate.An experimental analysis of the speech emotion recognition model of the improved convolutional neural network was conducted.Experiments show that the error recognition rate of the improved speech emotion recognition algorithm is reduced by about 7% compared with the traditional algorithm.In addition,This paper also implements speech emotion recognition based on ARM(Advanced RISC Machines)service robots and designs a speech emotion recognition system.According to the unique characteristics of the embedded system,this paper implements the transplant of speech emotion recognition module.