Removal of artifact from EEG signal using differential evolution algorithm

Electroencephalogram (EEG) is the neurophysiologic measurement of the electrical action of the brain, acquired by recording from electrodes located on the scalp. EEG is a vital clinical tool for diagnosing, monitoring and managing neurological disorders. EEG signal is contaminated with various artifacts such as Electroocculogram (EOG), Electrocardiogram (ECG) and Electromyogram (EMG). In this paper, we propose a novel method called ANFIS-DE (Adaptive Neuro Fuzzy Inference System (ANFIS) tuned by Differential Evolution (DE) algorithm) to estimate the artifacts and to extract the EEG signal from stained EEG signal. Differential Evolution (DE) algorithm is used to find the optimum design parameters of ANFIS to achieve better performance and faster convergence with simpler structure. Quantitative analysis of Signal to Noise Ratio and Mean Square Error reveals that ANFIS parameters tuned with Differential Evolution algorithm (ANFIS-DE) outperforms the ANFIS with general hybrid learning algorithm.