基于生理信号的情感识别技术综述

Review of Emotion Recognition Based on Physiological Signals

情感识别是人机情感交互的前提,也是实现人机自然交互的关键。介绍了情感识别的重要性以及基于人脸表情图像与语音信号的情感识别技术研究现状;从生理学实验研究角度,阐述了心率、脑电、肌电、心电、皮电、呼吸生理信号与人类情感表现的相关性,也说明建立生理信号与情感的关系模型有充分的实验依据;描述了如何利用各种信号分析和处理方法来提取生理信号特征,并通过优化合成可用于情感分类器的特征向量;着重介绍了神经网络、Fisher线性判别函数、K最近邻和多层感知器等机器学习算法应用于情感分类器的研究成果,最后展望了基于生理信号情感识别技术的应用前景,并总结了它的不足之处。

Emotion recognition is the base of human-machine emotion interaction as well as the key technology to realize human-machine nature interaction. Firstly, the importance of emotion recognition and the research status of emotion recognition based on face expression image and speech signal are introduced. Secondly, it is also clarified that the correlation between physiological signals, including heart rate, brain wave, myoelectricity, electrocardio, etc., and emotion exists based on the results of physiological experiments. Thereby, it is feasible to model the relation between physiological signals and emotion. The methods to extract the features from physiological signals of human and to combine a vector using the extracted features are depicted. The emotional classifiers based on machine learning algorithms including artificial neural networks, Fisher linear discriminating function, K nearest neighbor algorithm and multi-level perception are also introduced. Finally, the prospect of application about emotional recognition technologies using the physiological signals is described, and the defect of the technology is summarized as well.