Nonlinear features based classification of active and resting states of human brain using EEG

Electroencephalography is most common noninvasive neuroimaging modality and it is widely used for measuring brain electrical signals. Measurement of electrical signals from the scalp requires high density electrodes and low noise amplifier. It is well known fact that neural activity increased with increasing the mental work e.g., IQ task in our case. In this paper, non-linear features have been used to classify the active and resting states of the human brain. We have used EEG acquired from 08 healthy participants during IQ task and resting conditions. Nonlinear feature e.g., Approximate entropy, sample entropy and Composite permutation entropy index (CPEI) have been computed from recorded EEG data. These nonlinear features were fed to the classifier and we are able to classify the active and rest conditions. Also for classification, SVM produced better results with 89.1% and 92.5% accuracy for eyes open (EO) vs IQ and eyes open (EO) vs eyes close (EC) conditions respectively as compared to other classifiers. Also results compared with linear features extraction methods.