基于异构脑电特征图空间注意力迁移学习的情感识别OA

Emotion recognition based on EEG(Electroencephalogram)signals is limited by the volume of EEG data and the specifications of the equipment electrodes.The transfer learning methods convert EEG signals into brain functional topographic maps through interpolation and utilize pre-trained models for feature extraction,but this meth-od has limitations in the extraction of emotional information.To improve the performance of transfer learning and the generalization ability of emotion recognition,this study proposes an emotion recognition method based on heterogene-ous EEG feature map using spatial attention transfer learning.A brain feature mapping network based on spatial at-tention mechanism and residual network is pre-trained using the brain functional topographic data interpolated by in-terpolation method.As an alternative to interpolation method,the spatial attention mechanism is combined with trans-fer learning model to capture emotion information more effectively in the training of emotion recognition model.The proposed method is verified on DEAP,HCI,SEED-IV and SEED-V emotion recognition databases,and the accura-cy is 54.8%,63.6%,32.8%and 25.9%,respectively.