CROSS-SUBJECT EEG EMOTION CLASSIFICATION ALGORITHM BASED ON MODEL TRANSFER
The distribution of EEG emotion signal varies greatly among different subjects,which leads to low classification recognition rate.Therefore,this paper proposes a cross-subject EEG sentiment classification algorithm based on model transfer.The initial model of the convolutional neural network was obtained by training the selected source domain data.The distributed similarity between the target domain and the source domain was maximized by iterative nearest point algorithm.A new convolutional neural network model was obtained by fine-tuning to identify the target domain data.The experimental results show that the proposed algorithm can realize the common use of the initial network model by different subjects,which greatly improves the efficiency of the model.The domain adaptation algorithm based on iterative nearest point makes the classification accuracy of EEG emotion model transfer over 90％,which provides a new idea for EEG emotion classification of different subjects.
EEG emotion signalModel transferIterate the nearest pointDomain adaptive
国家科技期刊平台
NSTL
万方数据
