Cerebral asymmetry representation learning based deep subdomain adaptation network for EEG-based emotion recognition.

Researchers

Journal

Modalities

Models

Abstract

Extracting discriminative spatial information from multiple electrodes is a crucial and challenging problem for Electroencephalogram (EEG)-based emotion recognition. Additionally, the domain shift caused by the individual differences would degrade the performance of cross-subject EEG classification.To deal with the problems, we propose the Cerebral Asymmetry Representation Learning based Deep Subdomain Adaptation Network (CARL-DSAN) to enhance the cross-subject EEG-based emotion recognition. Specifically, the CARL module is inspired by the neuroscience findings that asymmetrical activations of left and right brain hemispheres occur during cognitive and affective processes. In the CARL module, we introduce a novel two-step strategy for extracting discriminative features through intra-hemisphere spatial learning and asymmetry representation learning. Moreover, the transformer encoders within the CARL module could emphasize the contributive electrodes and electrode pairs. Subsequently, the DSAN module, known for its superior performance over global domain adaption, is adopted to mitigate domain shift and further improve the cross-subject performance by aligning relevant subdomains which share the same class samples.To validate the effectiveness of the CARL-DSAN, we conduct the subject-independent experiments on the DEAP database, achieving the accuracies of 68.67% and 67.11% for arousal and valence classification respectively, and corresponding accuracies of 67.70% and 67.18% on the MAHNOB-HCI database.The results demonstrate that CARL-DSAN could achieve outstanding cross-subject performance in both arousal and valence classification.© 2024 Institute of Physics and Engineering in Medicine.