Developing a Tunable Q-Factor Wavelet Transform Based Algorithm for Epileptic EEG Feature Extraction

Brain signals refer to electroencephalogram (EEG) data that contain the most important information in the human brain, which are non-stationary and nonlinear in nature. EEG signals are a mixture of sustained oscillation and non-oscillatory transients that are difficult to deal with by linear methods. This paper proposes a new technique based on a tunable Q-factor wavelet transform (TQWT) and statistical method (SM), denoted as TQWT-SM, to analyze epileptic EEG recordings. Firstly, EEG signals are decomposed into different sub-bands by the TWQT method, which is parameterized by its Q-factor and redundancy. This approach depends on the resonance of signals, instead of frequency or scales as the Fourier and wavelet transforms do. Secondly, each type of the sub-band vector is divided into n windows, and 10 statistical features from each window are extracted. Finally all the obtained statistical features are forwarded to a k nearest neighbor (k-NN) classifier to evaluate the performance of the proposed TQWT-SM method. The TQWT-SM features extraction method achieves good experimental results for the seven different epileptic EEG binary-categories by the k-NN classifier, in terms of accuracy (Acc), Matthew's correlation coefficient (MCC), and F score (F1). The outcomes of the proposed technique can assist the experts to detect epileptic seizures.