Time-recurrent HMM decision tree to generate alerts for heart-guard wearable computer

In this paper, we propose a time-recurrent decision tree in order to generate time bound alerts by the wearable computing system in cases where in the wearer is found to be in a medical state that deserves to be attended. This decision tree leads to generate alerts by picking up most likely state sequence, based on trained Hidden Markov Models (HMM) and acquired real-time signal. We present simulation results based on Physikalisch-Technische Bundesanstalt Database (PTBDB), available on physionet.org as it contains ECG as well as VCG data records from 294 subjects (healthy as well as having various heart diseases). The construction and learning of decision tree was carried out using Iterative Dichotomiser 3 (ID3) algorithm, which was found suitable for varying importance of different parameters for different classes of heart disease.