Evaluation of Blind Source Separation methods for noise reduction in BSPM recorded during exercise

摘要

The application of Blind Source Separation (BSS) is now an accepted technique in the feld of biosignal processing. The reduction of noise (biological and technical artifacts) in body surface potential mapping recorded during exercise is a frequent goal. Two BSS methods (FastICA and temporal decorrelation) for removing noise are compared with respect to the ability to isolate artifacts. Datasets from 13 subjects were studied. For quantification of the results the changes of the root-mean-square (RMS) before and after denoising were analyzed. The use of both BSS methods improved the signal-to-noise-ratios (SNR) in a range of −4 to −10 dB.