Arrhythmic ECG Signals Extraction by Blind Source Separation

Abstract

In this paper the blind source separation (BSS) algorithm based on the known nonorthogonal approximate joint diagonalization of several time-delayed covariance matrices and the time-varying lag calculation procedure based on the proposed two stage phase alloca-tion method is investigated by applying it to arrhythmic ECG signals with time-varying period mixed with white Gaussian noise or with chaotic Hénon map noise (at SNR=15 dB). Simulation results show that algorithm is able to extract arrhythmic ECG signals with time-varying period from noise by performing nonorthogonal approximate joint diagonalization of only two covariance matrices – one cova-riance matrix at zero lag and another – at a lag corresponding to the the variable period, which is calculated from phase signature from cycle-to-cycle of the ECG signal. Ill. 3, bibl. 11, tabl. 1 (in English; abstracts in English, Russian and Lithuanian).