A Single-Channel EEG Automatic Artifact Rejection Framework Based on Hybrid Approach

Dealing with the physiological artifacts is a challenge in the period of recording electroencephalography (EEG), especially with a restricted number of channels. EEG contaminated with the Electromyogram (EMG) and electrooculogram (EOG) can affect the subsequent analysis. In order to more effectively reject the artifacts and not degrade the quality of the recorded EEG signals, a new framework with signal decomposed technology and blind source separation (BSS) is proposed. First, the framework detects and classifies EMG and EOG artifacts, and then uses CEEMDAN-CCA and SSA-SOBI methods to remove artifacts for EMG and EOG signals respectively. The method of removing the artifacts of EOG combines fuzzy entropy to distinguish the components of EOG. The methods are evaluated on semi-simulated dataset. Almost all the artifact fragments can be detected and type of the artifacts can be classified correctly. The two denoising algorithms achieve high correlation with the original signal under a low SNR condition. The results demonstrate the superior performance in single-channel EEG automatic artifact rejection.