Wavelet-based denoising for EEG-based pattern recognition systems

Electroencephalogram (EEG) has been widely studied for EEG-based pattern recognition systems such as seizure, sleep stage, emotion, alcoholics and person recognitions. However, EEG signals are subject to noise and artifacts, which negatively affects to the pattern recognition systems. Hence, an effective EEG denoising technique is becoming necessary. In this paper, we propose an EEG denoising technique in which noisy signals are decomposed by a Wavelet transform operation, followed by Thresholding component using Energy Packing Efficiency, before being reconstructed to obtain the clean signals. The experiments are conducted on two EEG public datasets and the results show that our proposed technique achieves good performance on denoising EEG signals and improves EEG-based pattern recognition systems the most.