کنفرانس بین المللی کنترل ، ابزار دقیق و اتوماسیون

صفحه اصلی / هشتمین کنفرانس بین المللی کنترل ، ابزار دقیق و اتوماسیون

Application of NSGA-II in Channel Selection of Motor Imagery EEG Signals with Common Spatio-Spectral Patterns in BCI Systems

نویسندگان :

M. Moein Esfahani¹ Hossein Sadati²

1- K. N. Toosi University of Technology 2- K. N. Toosi University of Technology

کلمات کلیدی :

Brain-Computer-Interface, BCI, EEG, Motor Imagery, CSSP, Common Spatio-Spectral Pattern, Classification, NSGA-II

چکیده :

Today, Brain activities are the subject of research study in Brain-Computer Interface (BCI) systems as used in a wide variety of applications. Electroencephalography (EEG) signals as brain cognition methods are the most dominant ones for acquiring brain electrical activities. In order to obtain feature extractions in EEG signals with Motor Imagery (MI) tasks, the common spatio-spectral pattern (CSSP) has been proposed. CSSP is an effective method to discriminate and classify the EEG Signals based on movement Tasks in two-class motor imagery signals. It has been implemented and compared with conventional CSP, and its optimized methods have been described. In this study, meta-heuristic multi-objective non-dominated sorting GA algorithm (NSGA-II) has been performed to select an optimal subset of channels in the multi-channel EEG motor imagery signals. The goal is to find an optimal solution to the channel selection problem to select the best subset of channels in the multi-channel EEG signals in brain-computer interface (BCI) systems. The main goal of the channel selection procedure in EEG signal analysis is to achieve a limited number of channels so that the subject feels more comfortable with the gel-based EEG electrodes. In addition, this procedure reduces the problem of overfitting in the classification results in the presence of a large number of redundant channels. Finally, the results indicate that the classification accuracy of the proposed CSSP method with 10-fold cross-validation achieves a higher accuracy in comparison with the results from CSP, CSP-TSM and LRCSP method