中文摘要：

针对采用人脸图像进行情感识别的准确率低、存在个体差异性且不能满足应急处置需求的问题，文章提出了一种基于图像和生理信号的多模态特征融合情感识别方法。利用奇异值分解（singularvaluede—composition，SVD）方法和小波分解法分别对图像信息和生理信号进行特征提取，然后采用主成分分析（principalcomponentanalysis，PCA）方法对多模态特征进行降维融合，将反向传播（backpropagation，BP）神经网络作为分类器，对不同情感进行分类识别。情感诱发试验结果表明，该方法能有效提高情感识别的正确率。

英文摘要：

It is not accurate by using facial images for emotion recognition and cannot meet the requirements of disposal of the emergency or deal with the individual differences. A multimodal information fusion method for emotion recognition based on facial images and physiological signals is proposed. Firstly, image information and physiological signals are extracted through singular value decomposition（SVD） method and wavelet packet method respectively. Secondly, dimensionality reduction of signal characteristic is conducted through principal component analysis（PCA） method. Finally, back propagation（BP） neural network is taken as classifier to recognize the different emotions. The experimental results show that this method can improve the accuracy of emotion recognition.