A hybrid ICA-Hermite transform for removal of Ballistocardiogram from EEG

In this paper the problem of removing Ballistocardiogram (BCG) artifact from EEG signal is addressed. BCG removal is an important task in analysis of simultaneous EEG-fMRI data. We propose a new method by combining independent component analysis (ICA) and discrete Hermite transform (DHT) for this purpose. Discrete Hermite transform is a powerful technique which is able to model a signal with no assumption about its shape. This feature makes DHT an appropriate tool to be combined with ICA for removing the BCG artifact. We show that the proposed hybrid ICA-Hermite transform can compensate for the existing drawbacks of the two methods, when applied separately. A significant improvement over conventional methods is demonstrated with synthetic data, and supported by preliminary work with real EEG.