Robust compressed analysis using subspace-based dictionary for ECG telemonitoring systems

Abstract

To realize Electrocardiography (ECG) signals monitoring systems, compressive sensing (CS) is a new technique to reduce power of biosensors and data transmission. Instead of spending high complexity on reconstructing back to data domain to do signal analysis, compressed analysis (CA) exploits the data structure preserved by CS to directly analyze in the compressed domain. However, compressively-sensed signals contaminated by interference cause learning performance degradation. Meanwhile, traditional interference removal methods are developed for signals in data domain, which involve reconstruction. In this paper, we propose a new CA framework using pre-trained subspace-based dictionary to project interfered and compressed data onto the subspace with high learnability and low complexity. Through simulations, we show that our technique enables 5.64% improvements on accuracy of detection compared with conventional CA, and reduces 99% complexity compared with reconstructed analysis.