Combining Sub-band Power Features Extracted from Different Time Segments of EEG Trials

In this paper, we proposed an optimum combination of sub-band power features method for improving the classification accuracy rate of left-or right-hand movement imagery electroencephalogram signals. The sub-band power features were extracted from the best time segment of electroencephalogram trials and the proposed training model determined the optimum combination of sub-bands. Our approach was successfully applied to the BCI Competition 2003 Data Set III and achieved a classification accuracy rate of 92.9% on the test data. The performance showed that our method outperformed the existing other researchers' results by achieving the highest classification accuracy.