Self?training maximum classifier discrepancy for EEG emotion recognition

Even with an unprecedented breakthrough of deep learning in electroencephalography (EEG), collecting adequate labelled samples is a critical problem due to laborious and time-consuming labelling. Recent study proposed solve the limited label via domain adaptation methods. However, they mainly focus on reducing discrepancy without considering task-specific decision boundaries, which may lead feature distribution overmatching therefore make it hard match within large gap completely. A novel self-training maximum classifier method for EEG classification this study. The approach detects from new subject beyond support existing source subjects by maximising discrepancies between two classifiers' outputs. Besides, that uses unlabelled test data fully use knowledge further reduce proposed. Finally, 3D Cube incorporates spatial frequency information create input features Convolutional Neural Network (CNN) constructed. Extensive experiments SEED SEED-IV are conducted. experimental evaluations exhibit can effectively deal transfer problems achieve better performance.