A Zero Padding SVD Encoder to Compress Electrocardiogram

Abstract

A new method to compress electrocardiogram (ECG) signals, whose novelty is related to the choice of an appropriate basis of representation for each ECG to be compressed using the Singular Values Decomposition (SVD), is proposed in this paper. The proposed method, named Zero Padding SVD Encoder, consists of two steps: a preprocessing step where the ECG is separated into a set of signals, which are the beat pulses of the ECG; and a compression step where the SVD is applied to the set of beat pulses in order to find the basis that better represents the entire ECG. The elements of the basis are encoded using a wavelet procedure and the coefficientes of projection of the signal on the basis are quantized using an adaptive quantization procedure. Numerical experiments are performed with the electrocardiograms of the MIT-BIH database, demonstrating the efficiency of the proposed method.