Classification of mental tasks using S-transform based fractal features

Brain Computer Interface is a reliable communication interface between human brain and external world. It translates human brain electrical activity to useful command by extracting meaningful features from Electroencephalogram signals. In present work, feature extraction techniques and classification methods are proposed for implementation of Brain Computer Interface system. Proposed methodology is carried out in four methodological steps. At first step, segmentation and windowing of Electroencephalogram signals are performed. The S-transform of segmented Electroencephalogram signals is evaluated in second step. At third step, mean and maximum values of Katz's Fractal Dimension are calculated from S-transform coefficients as features. Classification of extracted features is carried out in the fourth step using three machine learning techniques viz. Random Forest, Artificial Neural Network and Support Vector Machine. Classification results reflect the efficiency of S-transform based feature extraction technique in Brain Computer Interface implementation.