A classification method of different motor imagery tasks based on fractal features for brain-machine interface

The objective of this study is to classify spontaneous electroencephalogram (EEG) signal on the basis of fractal concepts. Four motor imagery tasks (left hand movement, right hand movement, feet movement, and tongue movement) were investigated for each EEG recording session. Ten subjects volunteered to participate in this study. As we known, fractal geometry is a mathematical tool for dealing with complex systems like EEG signal. Therefore, we used the fractal dimension (FD) as feature for the application of brain-machine interface (BMI). Effective algorithm, namely, detrended fluctuation analysis (DFA) has been selected to estimate embedded FD values between relaxing and imaging states of the recorded EEG signal. To show the pattern of FDs, we propose a windowing-based method or also called time-dependent fractal dimension (TDFD) and the Kullback-Leibler (K-L) divergence. The K-L divergence and different expected values are employed as the input parameters of classifier. Finally, featured data are classified by a three-layer feed-forward neural network based on a simple backpropagation algorithm. Experimental results show that the proposed method is more effective than the conventional methods.