多模态情感识别综述

Overview of multimodal emotion recognition

情感是人对客观事物的态度体验及相应的行为反应,而情感识别则是人机交互中极其重要的问题。文章围绕多模态情感识别技术,分别梳理了文本、语音和图像等模态特征提取与情感分类的发展进程和研究现状,着重探讨了多模态情感识别的特征级融合、决策级融合、模型级融合方法,最后总结了现阶段多模态情感识别中存在的问题和未来的研究方向。

Emotion is the attitude,experience,and corresponding behavioral response of a person towards objective things,and emotion recognition is an extremely important issue in humancomputer interaction.The article focuses on multimodal emotion recognition technology,and summarizes the development process and research status of feature extraction and emotion classification for text,speech,and image modalities.It particularly explores feature level fusion,decision level fusion,and model level fusion methods for multimodal emotion recognition.Finally,the article summarizes the current problems and future research directions in multimodal emotion recognition.