Performance Analysis of Mother Wavelet Functions and Thresholding Methods for Denoising EEG Signals during Cognitive Tasks

Artifact reduction is a major challenging task in the analysis of Electroencephalography signals. Wavelet denoising is one of the efficient techniques for rejecting artifacts and it has been employed in the detection of various brain diseases. The selection of the most suitable wavelet function and the thresholding method, which varies with the nature of the task performed, are important steps in the process of wavelet denoising. In this study, the effects of wavelet denoising on Electroencephalograph (EEG) signals during reading and mental arithmetic tasks were investigated. The effectiveness of each of the 46 wavelet functions including Daubechies- db, Symlet- sym, Coiflet- coif and Discrete Meyer- dmey, was examined for denoising the EEG signals recorded during reading and mental arithmetic tasks. EEGs of ten healthy subjects were recorded during reading and mental arithmetic tasks. The performance evaluation of wavelet function and thresholding method were done on the basis of peak signal to noise ratio (PSNR), root mean square error (RMSE), percentage root square difference (PRD), and cross correlation coefficient (CC). Subsequently, the EEG signals were decomposed into different frequency bands, and variations in EEG band power were computed before and after denoising. The wavelet function "dmey" was found to be the most efficient wavelet function for denoising EEG signals during cognitive tasks. Rigrsure thresholding method was found to be the best thresholding method in this study. A significant difference in EEG rhythmic activity (P <; 0.05) was observed after denoising the EEG signal.