EEG Channel Selection Algorithm Based on Attention Mechanism for Emotion Recognition
作者
Yifan Bai,Liying Yang,Ruiqi Gao,Qiang Wang,Jiayun Song,Kai Yao
出处
期刊: [Institute of Electrical and Electronics Engineers] 日期:2025-12-15卷期号:: 6950-6957
标识
DOI:10.1109/bibm66473.2025.11356915
摘要
Constructing emotion recognition models based on electroencephalogram (EEG) signals is crucial for implementing practical brain-computer interface applications. However, EEG signals are typically recorded using numerous non-invasive electrodes through wearable devices, which increases computational complexity and reduces user comfort. Effective channel selection algorithms can reduce the number of required electrodes and decrease device complexity and size, thereby improving wearability. Attention mechanisms, analogous to human cognitive processes, can dynamically focus on the most important information. In EEG-based emotion recognition, attention mechanisms can dynamically select the most relevant electrode channels that contain emotional state information. This paper proposes the Global Second-order Pooling Channel and Multi-head Temporal Encoding Network (GCMTE-Net), an attention mechanism-based EEG channel selection algorithm. The proposed method integrates a channel attention mechanism utilizing Global Second-order Pooling (GSoP) techniques with multi-head temporal encoding to achieve effective EEG feature extraction and channel selection. Experiments conducted on the DEAP dataset demonstrate that the model achieves 98.6% accuracy using only 24 selected channels. Channel suppression experiments further validate the superiority of the proposed channel selection approach.