Motor Imagery Features Selection with Grey Wolf Optimization for Brain Computer Interface

Feature selection is a significant process in brain-computer interfaces (BCIs) that has the potential to significantly enhance system performance, especially if a large number of features are present. Feature selection can help decrease noisy data, overtraining effects, storage needs, and computational complexity thereby enhancing classifier performance. The various feature selection strategies have been reported to attain these objectives. In this work, a hybrid feature selection technique is suggested. It uses a wavelet packet decomposition technique to extract features and grey wolf optimization (GWO) technique to obtain the best subset of features. It is then evaluated with support vector machine (SVM) classifier and obtains maximum classification accuracy of 82.88% for the approximate entropy feature. Experimental findings reveal that the suggested approach outperforms the contemporary MI-BCI systems even with relatively small training data sets.