کنفرانس مهندسی برق ایران

صفحه اصلی / سی و سومین کنفرانس بین المللی مهندسی برق

Comparison of Channel Selection Methods for EEG Signal Classification

نویسندگان :

Soraya Charkas¹ MohammadBagher Shamsollahi²

1- دانشگاه صنعتی شریف 2- دانشگاه صنعتی شریف

کلمات کلیدی :

EEG،Channel Selection،Classification،Correlation،Granger Causality،Random Forest

چکیده :

This paper presents a comparative study of two major paradigms for EEG channel selection: feature-based selection and connectivity-based selection. The feature-based selection methods, including correlation-based and random forest-based approaches, prioritize channels based on their contribution to classification accuracy through feature extraction and ranking. On the other hand, the connectivity-based selection approach leverages Granger causality to identify channels based on their causal relationships and functional connectivity within the EEG signals. The comparative performance of these methods is evaluated using a common EEG dataset for classification tasks. Experimental results demonstrate that the random forest-based method, which ranks channels based on feature importance, outperforms both the correlation-based method and the Granger causality-based method in terms of classification accuracy. This highlights the effectiveness of feature extraction and ranking techniques for EEG channel selection in machine learningdriven EEG signal classification. However, the connectivitybased approach provides additional insights into the functional relationships between EEG channels, which can be valuable for understanding brain dynamics.