Automatic EEG artifact removal by independent component analysis using critical EEG rhythms

EEG is the recording of electrical activity in the brain. Four major frequency ranges identified in EEG are called delta, theta, alpha and beta rhythms. A system for removal of artifact from multichannel EEG signal is designed based on Independent Component Analysis using rhythmic components. Since different artifacts are overlapped with different EEG rhythms wavelet decomposition as a preprocessing step increases the redundancy and rejection of the suitable wavelet components decreases the probability of rejecting useful information. Two step identification of artifactual content before and after ICA using Kurtosis as a parameter helps in automation of the artifact removal process. This automation increases the accuracy and reduces the processing time. Using this method we can process more number of channels easily. The wavelet coefficients can be easily processed and therefore these transforms can be used as a preprocessor in clinical applications where various states of EEG signal correspond to different representation of information.