Personalized Natural Language Emotion Recognition System Based on Hybrid Neural Network
In order to improve the robustness and generalization ability of the system when there are slight changes or noise interference in the input natural language information,a personalized natural language emotion recognition sys-tem based on hybrid neural network is designed.Taking field programmable gate array as the core,a natural language acquisition module is designed to collect personalized natural language information.Processing the voice information in the collected personalized natural language information through a preprocessing module to obtain a logarithmic Mel spectrogram.Emotion recognition module uses dynamic convolution neural network and long-term and short-term memory network to build a hybrid neural network,in which,through dynamic convolution neural network,voice infor-mation features are extracted from logarithmic Mel spectrogram,and through long-term and short-term memory net-work,text information features are extracted from natural language text information,and through fully connected neural network,voice and text information features are fused to output personalized emotion recognition results.Experiments show that the system can effectively collect personalized natural language information,extract phonetic information fea-tures and complete personalized natural language emotion recognition.Under the interference of noise,the emotion recognition accuracy of the system is high.
hybrid neural networkpersonalizationnatural languageemotion recognitiondynamic convolutionlong-term and short-term memory network
万方数据
维普
