Hierarchical fusion of visual and physiological signals for emotion recognition

Emotion recognition is an attractive and essential topic in image signal processing. In this paper, we propose a multi-level fusion method to combine visual information physiological signals for emotion recognition. For information, serial of two-stage features enhance the representation facial expression video sequence. We integrate Neural Aggregation Network with Convolutional feature map reinforce vital emotional frames. signals, parallel scheme widen band annotation electroencephalogram signals. extract frequency Linear-Frequency Cepstral Coefficients it complexity denoted by Sample Entropy (SampEn). classification stage, realize both level decision information. Experimental results validate effectiveness proposed multi-modal method.