Wavelet domain blind signal separation to analyze supraventricular arrhythmias from holter registers

Abstract

Detection of atrial activity (AA) is quite important in the study and monitoring of supraventricular arrhythmias. This study shows the possibility of AA extraction from atrial fibrillation (AF) episodes in Holter registers using only two leads with a new technique, the Wavelet Domain in Blind Source Separation (WDBSS). Our principal aim is to join a processing stage with Blind Source Separation (BSS) with methodologies based on wavelet transform. A first stage with Discrete Wavelet Transform (DWT) increases the spectral information, decomposing the considered signal in a set of coefficients with different temporal and spectral features. A second stage with BSS uses this information to extract the AA. The obtained improvements are the increase of spectral concentration (in the band of 5-8 Hz) and the lack of residual complexes. In WDBSS, the use of several leads from the ECG is needless, which could be applied for to the detection of different arrhythmias in Holter registers, where the number of leads is reduced, like the paroxysmal atrial fibrillation. © Springer-Verlag 2004.