Efficient Removal of Artifacts from EEG Recordings by ICA and Fast ICA Techniques

EEG, a highly complex signal, is one of the most common sources of information used to study brain function and neurological disorders. A vital problem in EEG analysis is the mixing of the signals with various other signals, such as eye blinks or pulse signals. These signals must be removed for precise analysis of EEG. One main challenge for EEG denoising is the identification of the artifactual components in EEG. ICA is a statistical method by which corrupted EEG signal is decomposed and artifacts in it can be detected. This paper illustrates the ICA and fast ICA technique for EEG artifact detection and removal. Before applying ICA, preprocessing of the EEG data is done using Centering and Sphering. From results we found that, time needed to separate the EEG signals from mixture of signals using ICA algorithm is very large. Therefore, Fast ICA algorithm is developed and results shows that time required to separate the components using our fast ICA algorithm is very less as compared to ICA. Almost same accuracy as in ICA is achieved by our Fast ICA technique, with some time improvement.