A Hidden Markov Model-based approach to removing EEG artifact

Non-invasive methods of brain activity such as the Electroencephalogram (EEG) is a popular method which used in development of biomedical devices as an interface for many application like diagnostic, scientific, therapeutic and restorative. EEG recordings are usually contaminated with physiological and non-physiological artifacts. Major physiological source for EEG contamination is ocular artifacts, specially in a closed-loop and real-time application like brain computer interface (BCI), which it require some signal processing techniques for removing it. Several different method have been used to remove these artifact from EEG signals, but it is still a challenge to present clear data from contaminated signals. So, this paper proposed a method based on Hidden Markov Model (HMM) for removing artifacts which caused by eye blinks. In the HMM model, Baum-Welch procedure has been used to train network, which after detecting the eye blinks, it will be replaced by appropriate signals from transition probability. The estimated eye blink artifacts are evaluated by original signals, and the results shown outstanding performance of proposed method comparing to conventional EEG artifact removal method.