A Dual-PSoC based reconfigurable wearable computing framework for ECG monitoring

We propose a reconfigurable framework using Dual-PSoC (Programmable System-on-Chip) based wearable computing system capable of ECG monitoring. Keeping in view wearability constraints involving comfort, ease and usability, we chose to use vectorcardiographic leads for minimizing the number of electrodes. It also helped to choose proper parameters to capture entire cardiac activity instead of relying only on Heart Rate Variability (HRV). Hidden Markov Model (HMM) combined with Timed Automata (TA) was used for modeling the ECG signal and extracting the features. A combination of Viterbi algorithm and Time-recurrent decision trees were used for classification of the input ECG signal and alert generation in case of abnormality noticed beyond threshold level. In all eight class models of heart diseases have been defined. Physikalisch-Technische Bundesanstalt Database (PTBDB) available at physionet.org was used for modeling. Data specific to these HMM models as well as decision trees were stored in the form of lookup tables in PSoC memory. As a part of the proposal, a belt embedded data acquisition unit has been envisioned for providing signals acquired from electrodes to PSoC. PSoC with wireless module was used for signal processing and further analysis in terms of classification and possible alert generation. The scheme envisages use of two PSoCs such that while first PSoc plays a role in data acquisition and processing, second one shoulders the responsibility of classification and decision making.