Hand Movement Imagery Task Classification using Fractal Dimension Feature

Abstract

In this paper, a nonstimulus-based Brain Machine Interface (BMI) approach is used to acquire the brain signal from ten different subjects using 19 channel EEG electrodes while performing four different hand movement imaginary tasks. Three different Fractal Dimension algorithm namely Box counting algorithm, Higuchi algorithm, and Detrended fluctuation algorithm are used to extract fractal dimension features from recorded EEG signal and associated with the respective mental tasks. Three Feed-Forward Neural Network model is developed. The performance of the three Neural Network model is evaluated in term of classification rate and compared. The performance of the developed network models are evaluated through simulation. It is observed that the neural network model trained with Higuchi algorithm has contributed high classification accuracy with the better training and testing time for all 10 subjects. The result clearly indicates that the Higuchi fractal dimension algorithm can be used as a feature to classify motor imagery task for the proposed BMI system.