Application of SVD for Removing Motion Artifacts from the Measurements of a Wireless Electrocardiogram

Cardiovascular diseases (CVD) claim tens of millions of lives worldwide every year. About one-third of these die before they reach 70. For decades, a considerable effort has been made to supplement clinical treatments with telemedicine. In this respect, wireless electrocardiograms play a vital role, since affordable, unobtrusive, and long-term monitoring can be made with them while patients carry out everyday activities unhindered. Moreover, symptoms which can otherwise be hidden during short-term, clinical check-ups can be detected and exact causes can be assigned to them. Nevertheless, wireless electrocardiograms are highly sensitive to motion. Even though hardware and software solutions have been proposed in the past to remove motion artefacts, the results are still unreliable. In this paper we propose (1) to use inertial sensors to directly measure the motions affecting the electrodes of a wireless electrocardiogram and to correlate these measurements with motion artefacts and (2) to employ a dimensionality reduction technique (singular value decomposition, or, in short, SVD) in order to recover the underlying useful ECG signals. We consider different types of intense movements and confirm that SVD consistently and reliably enables to reconstruct the QRS complex and to some extent the T waves. SVD, however, is unable to recover the P and T waves in some irregular and complex motions.