Correction of ocular artifacts from single channel EEG by EMD with an adaptive thresholding

Electroencephalogram (EEG) is a widely used signal for analyzing the activities of brain. It has extensively used for the diagnosis of different nervous system disorders such as Alzheimer's, Parkinson's, Seizures, Epilepsy, etc. Ocular activity creates significant artifacts in electroencephalogram recordings. These artifacts increase complexity in analyzing the EEG and obtaining the clinical information. This paper proposes wavelet thresholding principle to the IMF resulting from applying EMD to a signal. This technique is assessed on EEG signals taken from polysomnographic records. Performance metrics such as change in Signal to Noise Ratio (ΔSNR), Artifact Rejection Ratio (ARR), and Normalized Mean Square Error (NMSE) have been evaluated for measuring efficiency of each of the thresholds in DWT, EMD Interval Thresholding(EMDIT) and Iterative EMD Interval Threshold (IEMDIT) techniques respectively. Result of this study reveals that IEMDI+PT and IEMD+PT has shown superior performance in terms of ΔSNR, ARR, NMSE and effectively eliminates the artifacts from single channel EEG signals.