Selection of proper electrodes and improving performance in brain computer interfaces

This study contains mental and motor imagery experiments with 6 different subjects, in order to see the effect of using few electrode channels with an efficient feature extraction algorithm for an online brain computer interface application. Independent Component Analysis (ICA) and Continuous Wavelet Transform (CWT) methods are compared for their discrimination ability. The electrode channels which define different regions of the brain evaluated separately and the their classification performances are given by well-known classifiers. While the best classification performance with ICA is on frontal (F3-F4) region with 87%-85%, with CWT it has close performance values for frontal (F3-F4), central (C3-C4) and occipital (O1-O2) regions as 86%, %86 and 88%. O1 and F3 channels have the highest performance. The total classification time for CWT with Neural Networks is 100 seconds and 138 seconds for ICA. Therefore, it can be concluded that CWT can be a successful feature extraction method for online brain computer interface applications which contains imagery mental and motor tasks.