Removal of EOG and EMG artifacts from EEG signals using blind source separation methods

Artifacts in EEG signals may be referred to as a disturbance in the EEG recordings that does not derive from the brain. EEG recording is in the millivolt range, and it includes the electrical activity from the muscles under the scalp. This recording often includes the artifacts due to power source, electrode movement, eyes movement, muscle movements, electrode pop, and sweating. This paper mainly concentrates on the artifacts which occurs during the eyes and muscle movement. In the proposed work, different BSS methods like FastICA, WASOBI, and EFICA algorithms are used to compare the reduction of artifacts in terms of MSE and PSNR in EEG signal. Different combinations using FastICA, WASOBI, and EFICA are evaluated and compared their performance to find the best COMBI method. Among the COMBI methods, the EFICA-WASOBI combination provides low MSE and high PSNR value, which indicates the high artifact reduction. In EFICA-WASOBI, the average value of MSE and PSNR are 0.021 and 42.78dB respectively for EMG artifact reduction and 0.060 and 32.97dB for EOG artifacts reduction using EEGdenoiseNet dataset. Similarly, the average MSE value is 0.018 and average PSNR value is 49.19dB for EOG artifact reduction using Mendeley dataset. The results show that by using the combination of EFICA-WASOBI, the effect of EOG and EMG artifacts can be reduced.