کنفرانس بین المللی محاسبات نرم

صفحه اصلی / ششمین کنفرانس بین المللی محاسبات نرم

Temporal–Spatial Graph- Integrated Framework for EEG- Based Emotion Recognition Using LSTM–GCN Architecture

نویسندگان :

Zahra Amiri¹ Abdorreza Hesam Mohseni²

1- دانشگاه گیلان 2- دانشگاه شریف

کلمات کلیدی :

EEG،Emotion Recognition،LSTM،GCN،SEED،Deep Learning

چکیده :

This paper presents an advanced framework for EEG- based emotion recognition that synergistically combines temporal and spatial neural modeling. A dual- stage architecture is employed, where Long Short- Term Memory (LSTM) networks extract dynamic temporal features from multi- channel EEG signals, while Graph Convolutional Networks (GCN) capture spatial correlations among electrodes through a structured brain topology. The approach leverages both temporal evolution and spatial connectivity to learn discriminative representations of affective states. Evaluations on a benchmark dataset confirm that the proposed LSTM- GCN fusion model effectively enhances the robustness and generalization of emotion classification compared to conventional deep learning baselines. The findings highlight the potential of graph- aware temporal modeling for next- generation affective computing applications. In this study, we successfully achieved promising results, including an accuracy of 97.98% and a loss of 0.0558.