Reconstruction for ECG Compressed Sensing Using a Time-Normalized PCA Dictionary

Abstract

Compressed sensing (CS), due to its computational simplicity is a perspective data reduction technique for remote ECG monitoring applications. In this paper, a novel method of reconstruction for CS of ECG signal is proposed, which uses a time-normalized agnostic dictionary created by the principal component analysis (PCA) of training signals. The proposed method exploits a QRS detector to split the input signal into variable-size frames and shows significantly better reconstruction quality compared against traditional orthogonal matching pursuit (OMP) approach with Mexican hat and Symlet4 wavelet dictionaries.