Intracardiac ECG pulse localization using overlapping block sparse reconstruction

Abstract

The processing and analysis of multichannel signals which exhibit some synchronicity is of critical importance in biomedical signal processing. Though sparse reconstruction of such signals is well known, penalties enforcing both sparsity and partial synchronicity between channels have been less investigated. In this paper, we present an algorithm based on overlapping grouped LASSO to guarantee both sparsity and synchronicity for multichannel signals. We show that the proposed method can be implemented efficiently, using consensus learning approaches and recent work of the authors for accelerating the related optimization procedure. Results on synthetic data show the efficiency of the proposed approach. When applied on real intracardiac recordings, the proposed method succeeds in automatically detecting electrical pulses both for sinus rhythm and atrial fibrillation episodes.