Evaluation of PSE, STFT and probability coefficients for classifying two directions from EEG using radial basis function

Abstract

EEG (Electroencephalography) is a recording of electrical activities of brain measured from scalp. Brain is a control center for almost all functions of body. As EEG originates from brain, it contains various components related to cognitive activities of brain. Hence, it also contains information regarding the motor functions associated with movement of the body. EEG is commonly recorded for purposes of diagnosis and research associated with diseases like epilepsy, seizures, sleep disorders etc. But apart from these applications it can also be used to map various motor movements being thought of. This may lead to development of landmark devices in the field of rehabilitation of physically challenged individuals. Here we intend to extract the features and classify the directions using EEG. At initial stage it is desired to classify two movements i.e. left and right, but the method can be extended for the classification of other directions as well. In present scenario the most suitable methods for classification problems can be developed using machine learning algorithms. In this work the features like probability co efficient, PSE (power spectral entropy) and STFT (Short Time Fourier Transform) are extracted and evaluated for their efficiency in classification. Radial Basis Function is used for classifying these features. The study shows probability co efficient and STFT have yielded about 60% accuracy in classifying raw EEG signals proving them advantageous over power spectral entropy.