基于语音信号与文本信息的双模态情感识别

Multimodal Emotion Recognition Based on Speech Signal and Text Information

情感识别已成为人机交互不可或缺的部分,目前单模态情感识别具有识别率低、可靠性差的特点,故提出一种融合语音信号与文本信息的双模态情感识别方法。首先,采集特定情感状态下的语音信号及文本信息;然后提取语音相关特征参数以及文本情感关键词特征参数并对其进行优化;最后,对两个单模态识别器的输出结果进行加权融合获得识别结果。针对所提算法进行了相关实验研究,结果表明双模态情感识别技术具有更高识别精度。

Emotion recognition has become an indispensable part of human-computer interaction. This paper propsesa fusion method of speech signal and the text information in emotion recognition,because of the low recognition rate and poor reliability of single modal emotion recognition. First of all,collecting specific emotional state of the speech signal and text information;then extracting the speech feature parameters and keywords emotional char acteristic parameters of text information and optimize it; finally, recognition results are obtained by weighted fusion of the output results of two single modal identification devices. According to the results of experimaental,it showed that the dualmodal emtoion recognition technology has higher recognition accuracy.