The research of emotion recognition based on multi-source physiological signals with data fusion

Abstract

In the context of the increasing application value of emotion recognition and the continuous development of data fusion technology, it is of great significance to study the emotion recognition model based on multi-source physiological signals with data fusion. In this paper, the onedimensional-convolutional neural network-support vector machine (1D-CNN-SVM) emotion recognition model is constructed to extract the emotional features of multi-source physiological signal data, realize data fusion and complete emotion recognition. Firstly, based on the data level fusion method, dimension splicing for data of each channel is used to compare and analyze different data splicing combinations to explore the best one. Secondly, based on the feature level fusion method, the depth features of each part are fused and extracted by convolutional neural network models. Finally, feature stitching and support vector machine algorithm are used to classify and recognize emotion categories. The experimental results verify the effectiveness of the proposed model in the valence-arousal of the four-class task on DEAP dataset, and the recognition accuracy of the optimal combination can reach 93.10%.