Abstract:
This thesis proposes a novel approach for selecting the optimal subset of EEG channels that can
best discriminate between different types of events. The proposed method combines common
spatial pattern (CSP) filtering, artifact removal, linear discriminant analysis (LDA) classification,
and genetic algorithm (GA) optimization. The proposed methodology involves several steps,
including preprocessing, event separation, artifact removal, initial population creation, fitness
calculation, genetic algorithm iteration, and validation. The aim is to find the best set of EEG
channels that can best discriminate between different types of events using a combination of
machine learning and genetic algorithm techniques. The proposed approach has the potential to
enhance the accuracy and reliability of EEG signal analysis, while considering the individual
variability, spatial resolution, and frequency band of the signal. The approach is tested on a
publicly available dataset BCI Competition IV Datasets 2a and 2b, and the results show that it
outperforms existing methods in the literature with the accuracy of 87.40 % at 16 channels.
Overall, the proposed methodology has important implications for advancing the state-of-the-art
in EEG signal channel selection and improving the interpretability and clinical utility of EEGbased brain-computer interfaces.
