An Automated Motion Artifact Removal Algorithm in Electrocardiogram Based on Independent Component Analysis

摘要

Mobile ECG recordings are widely used to monitor abnormality of the cardiovascular system during daily life. However, ambulatory ECG recordings are often contaminated by many types of artifacts. In motion artifacts removal, because ECG and movement are not always independent, ICA based noise reduction may distort the signals. Thus, this paper introduces automatic noise detection and removal technique based on independent component analysis (ICA) preventing the signal distortion and attenuation. Using several proposed decision rules, 3-channel ECGs are analyzed their noisiness and Gaussianity to predict whether motion artifacts and ECGs can be separated and reconstructed without distortion. This method is evaluated by ECGs recorded during 0 to 7km/h rest, walking, and running exercise. Finally, its performance is compared to the conventional approaches of ICA-based noise reduction in ECGs. As results, the reconstructed ECGs by the proposed algorithm show higher correlation with estimated reference signals and there is no distortion in reconstruction of the signals.