Research of Discrimination Between Left and Right Hand Motor Imagery EEG Patterns Based on Tunable Q-Factor Wavelet Transform

In view of the problem of low accuracy and mutual information in left and right hand motor imagery-based ElectroEncephaloGram (EEG), a new approach based on Tunable Q-factor Wavelet Transform (TQWT) is proposed to handle with the binary-class motor imagery EEGs. Firstly, the TQWT is utilized to decompose the filtered EEG signal. Then, several sub-band signals are extracted and followed by calculating their energy, AutoRegressive (AR) model coefficients and fractal dimension. Finally, a Linear Discriminant Analysis (LDA) classifier is used to classify these EEGs. Two Graz datasets of BCI Competition 2003 and 2005 are employed to verify the proposed method. The maximum accuracy of classifying EEGs of four subjects is 88.11%, 89.33%, 77.13% and 78.80%, respectively, and the maximum mutual information is 0.95, 0.96, 0.43 and 0.45. The high accuracies and mutual information demonstrate eventually the effectiveness of the proposed method.