A fast algorithm for automated independent process separation from single channel biomedical signal recordings: FastIPA

Independent component analysis (ICA) has found many uses in source separation in biomedical signals. We highlight a methodology and put forward an algorithm which allows single channel ICA to be performed on single channel biomedical signal recordings. The algorithm uses a fast, deflationary approach to efficiently extract independent processes underlying the single channel recordings. We show that for processes which are reasonably spectrally disjoint the algorithm can separate out individual sources. We show examples of this using brain signal recordings and abdominal foetal recordings.