Chinese Speech Emotion Recognition System Based on LSTM

• 1、School of Computer, Bejing University of Posts and Telecommunications

Abstract：Emotions are one of the most important ways for human beings to express themselves, affecting people\'s daily work, life and important decisions. With the continuous development of computer technology, speech emotion recognition technology has also received extensive development and attention. In this paper, based on the background of a voice work report system of an enterprise, a voice emotion recognition algorithm that integrates multiple features and deep learning is proposed. The algorithm uses OpenSmile tool to extract various features of Chinese speech, and uses long and short-term memory network to obtain speech emotion, and achieves an accuracy of 87.2% on the CASIA Chinese speech emotion dataset. At the same time, based on this algorithm, this paper designs and implements a Chinese speech emotion recognition system, which has the functions of emotion labeling and emotion recognition, and has practical application value.

Keywords： computer software emotion recognition emotionq annotation long short term memory network multimodal fusion