Discriminating the brain activities for brain–computer interface applications through the optimal allocation-based approach

The translation of brain activities into signals in brain–computer interface (BCI) systems requires a robust and accurate classification to develop a communication system for motor disabled people. In BCIs, motor imagery (MI) tasks generate brain activities, which are generally measured by electroencephalogram (EEG) signals. The aim of this research was to introduce a method for the extraction of discriminatory information from the MI-based EEG signals for BCI applications. The proposed scheme develops an optimal allocation (OA)-based approach to discover the most effective representatives with minimal variability from a large number of MI-based EEG data. To investigate a suitable classifier for the OA-based features, the least square support vector machine (LS-SVM) and Naive Bayes (NB) methods are applied separately on the extracted features for discriminating the MI activities. Experimental results on datasets, IVa and IVb of BCI Competition III, show that the OA-based features with the LS-SVM classifier yields better performances compared to the NB classifiers. The results also demonstrate that the classification performance is improved up to 21.16 %, through the use of the OA algorithm with the LS-SVM, compared to the reported four prominent methods. This implies that the OA technique is promising for extracting representative characteristics from MI-based EEG data, which can be reliably used with the LS-SVM to identify different signals of brain activities in the development of BCI systems.