EEG-Based Emotion Classification Using Joint Adaptation Networks

Emotion classification based on EEG Signals are being increasing studied because of its applicability in human- machine interaction. However, in previous research, it is commonly assumed that the training and testing data share the same distribution. Unfortunately, this assumption is not always reasonable, for the variation of EEG can cause a substantial mismatch between datasets easily. The problem mentioned above results in degeneration of traditional emotion classification methods. In this paper, we construct a novel joint adaptation networks (JAN) to address this problem for emotion classification based on EEG. Experimental results on two representative EEG datasets demonstrate its validity. Moreover, further comparisons with the state-of-the-arts methods are also made to confirm its superiority.