Enhancing Motor Imagery EEG Classification Accuracy Using Weight Features Function

Brain-Computer Interface (BCI) is a computerized system that gathers, analyzes, and translates neural signals into commands, which are then transmitted to an output device to perform certain tasks. One of the most difficult parts of the BCI Motor Imagery-Electroencephalogram (MI-EEG) based system is the Classification Accuracy (CA). In order to get accurate classification, efficient and rapid features extraction is required for developing a successful MI-EEG classification model. In this article, the Motor Imagery (MI) of Left-Hand (LH) and Right-Hand (RH) actions is recognized using the Weight Features Function (WFF) that transforms initial features into more discriminant features to feed a Support Vector Machine (SVM) classifier. Appropriate weights were chosen by the Genetic Algorithm (GA) optimization method. Applying optimized WFF to four different datasets (IIIb from BCI competition III, III from BCI competition II, 2b from BCI competition IV, and Open Brain-Machine Interface (OpenBMI) dataset) made significant improvements in the CA for all studied datasets. Before using the WFF technique, the initial CA for the four datasets was 90.1%, 95.71%, 86.73%, and 83.83%. After applying the WFF technique, the CA is improved and achieves 96.1%, 100%, 94.2%, and 88.70% respectively