ABSTRACT
Brain-computer interface (BCI) has been widely introduced in many medical applications. One of the main challenges in BCI is to run the signal processing algorithms in real-time which is challenging and usually comes with high processing unit costs. BCIs based on motor imagery task are introduced for severe neurological diseases especially locked-in patients. A common concept is to detect one's movement intention and use it to control external devices such as wheelchair or rehabilitation devices. In real-time BCI, running the signal processing algorithms might not always be possible due to the complexity of the algorithms. Moreover, the speed of the affordable computational units is not usually enough for those applications. This study evaluated a range of feature extraction methods which are commonly used for such real-time BCI applications. Electroencephalogram (EEG) and Electrooculogram (EOG) data available through IEEE Brain Initiative repository was used to investigate the performance of different feature extraction methods including template matching, statistical moments, selective bandpower, and fast Fourier transform (FFT) power spectrum. The support vector machine (SVM) was used for classification. The result indicates that there is not a significant difference when utilizing different feature extraction methods in terms of movement prediction although there is a vast difference in the computational time needed to extract these features. The results suggest that computational time could be considered as the primary parameter when choosing the feature extraction methods as there is no significant difference between the results when different features extraction methods are used.
