Hidden Markov Models for Classification of Heart Rate Variability in RR Time Series

摘要

Discrete hidden Markov models (HMM) are trained in order to differentiate between RR series of persons in normal sinus rhythm and of post-myocardial infarct patients with six or more ventricular premature complexes per hour during 24 hour Holter ECG recording.The RR series of ten subjects of each group are used to train the models. 44 RR series of each group are used as test data. The basic RR intervals are transformed into the final observation sequences by means of symbolic dynamics.Models with different parameters of the symbolic dynamic building process are tested as well as different lengths of RR series and different numbers of hidden states of the model.