Recognition of motor imagery based on dynamic features of electroencephalography signals

摘要: The control of artificial limbs can be done by distinguishing the patterns of imagined movement using the Electroencephalography (EEG) signals. The aim of this study was to identify hand and foot imagery movements based on EEG signals. The IVA dataset of BCI Competition III, which includes EEG signals from 5 healthy individuals in C3, C4 and CZ channels, was used to design the imagery movements detection system. Initially, the basic components of EEG signal noise were removed using the MSPCA method. In the next step, the EEG signals were decomposed in two different ways including frequency filtering using the Butterworth filter and the wavelet packet transform (WPT). In this study, the detrended Fluctuation analysis, Fractal dimension, Correlation dimension, Lempel-ziv complexity and Entropy as nonlinear dynamics features, were calculated for the signals. In both decomposition methods, the desired features were calculated for the temporal version of the specified subbands. In order to determine the best performance of the system, different combinations of the channels and the features were evaluated. The wavelet-based decomposition method, in the case of using all three channels and five features, provided the highest recognition accuracy; So that using support vector machine (SVM) classification method, the accuracy of 93% was obtained in identifying the desired movements.