Statistical models based ECG classification
Résumé
This chapter gives a comprehensible description of two statistical approaches successfully employed to the problem of beat modeling and classification: hidden Markov models and hidden Markov trees . The HMM is a stochastic state machine which models a beat sequence as a cyclostationary Markovian process. It offers the advantage of performing both beat modeling and classification through a unique statistical approach. The HMT exploits the persistence property of the wavelet transform by associating to each wavelet coefficient a state and the states are connected across scales to form a probabilistic graph. This method can also be used for signal segmentation
Ce chapitre traite des aspects de reconnaissance des formes appliquées à la classification d'ECG pathologiques (arythmies, ischémies cardiaques). Deux approches sont proposées, développées et evaluées: un modèle de sous-battements basé sur les HMM et une approche d'Arbre de Markov Cachés (HMT).