WAVELET AND HMM ASSOCIATION FOR ECG SEGMENTATION

摘要

This paper presents a multiscale Hidden Markov Model (HMM) to improve an automatic segmentation of an electrocardiographic signal (ECG). While the HMM describes the dynamical mean evolution of cardiac cycle, the use of wavelet analysis in association with the HMM leads to take into account local singularities and to obtain better segmentation results. This was tested on a learning base composed of 130 patients.