Constrained based tunable Q wavelet transform for efficient decomposition of EEG signals

Abstract

Electroencephalogram (EEG) signal decomposition by selecting veracious basis function is a tedious task. In this paper, constrained based tunable Q wavelet transform (CTQWT) is proposed for adaptive selection of the optimum tuning parameters (Q and r) to decompose the highly nonstationary EEG signals accurately. The fitness function of mean square error (MSE) is used as a constraint to minimize the reconstruction error of EEG signals. Transformative optimization algorithms (TOA) are used to evaluate the optimum Q and r for the decomposition of the EEG signals. The efficacy of the proposed method is evaluated by comparing the reconstruction error obtained with traditional TQWT and CTQWT using the multiclass sleep stages and two class focal EEG datasets.