The Extraction of Ventricular Late Potentials Based on Sparse Component Analysis

Abstract

Blind source separation technology was applied to separate and extract different components from observed electrocardiograph signals. The electrocardiograph signal with ventricular late potentials presents the sparseness. Two-step approach of sparse component analysis was used in extraction of ventricular late potentials. An improved method of estimating the mixing matrix was proposed. It can improve estimation precision of mixing matrix about the source signals whose amplitudes have large difference in all or particular frequency bands, which realizes the source signal separation nondestructively. Simultaneously, the estimated signals are output in order solving the order uncertainty of independent component analysis. It is proved that sparse component analysis greatly enhances extraction precision of ventricular late potentials.