Entropy and Energy Based Reconstruction of EEG Motor Imagery Signals with Tunable Q-Factor Wavelet Transform

Abstract

Brain-Computer Interface applications aim to bridge the gap between individuals with motor impairments and technological devices through mental activity. In line with this goal, motor imagery signals acquired via electroencephalogram devices, which offer high temporal resolution, have become one of the challenges addressed by these applications. In studies developed in the literature, these signals are typically used in their raw or denoised forms during the classification process. Despite existing performances, the pursuit of methods that can enhance classification accuracy in this field continues. Therefore, this study proposes the enhancement and resynthesis of subband signals obtained through the Tunable Q-Factor Wavelet Transform using a method referred to as the Information Coefficient. In classification tasks conducted following this enhancement, the signal obtained with parameters Q:3 and J:7 was observed to improve accuracy by 1.34%.