Adaptive EEG noise filtering for coherence analysis

In this paper, we propose an automatic electroencephalography (EEG) noise removal method for coherence analysis. EEG signal is generally contaminated with many noise sources such as eye movements, cardiac signals, muscle noise and line noise, etc. Before the analyzing, we must eliminate these noise components. There is a lot of noise removing technic. ICA is famous technic for effective noise removing. But, ICA is not suitable for coherence analysis due to the changing of electrode's spatial information. For general coherence analysis, the electrode's spatial information is important. However, the spatial information is vanishing after applied the ICA. So, general ICA approach is not suitable for the coherence analysis. In our study, we use ICA for getting each ICs. Each ICs are analyzed by exponential analysis for distinguishing noise ICs and normal ICs. After distinguishing noise and normal ICs, we used noise ICs for reference input of adaptive filter. The experimental results of the proposed method shows good performance.