Session S41.3

A Distributed Algorithm for Real Time Analysis of Heart Rate Variability Using Low Power Wireless Monitors

E Ertin*, L Wittmers, M al'Absi, S Kumar

Ohio State University
Columbus, OH, USA

Heart Rate Variability (HRV) analysis provides information on the autonomous nervous system activity, with links to disease conditions and mental and emotional states. We consider distributed computation of HRV spectra using a low power wireless ECG heart rate monitor that is streaming data to a mobile computing device such as a cellular telephone. The distributed system enables real time, long term ambulatory monitoring of HRV spectra. The wireless ECG heart rate monitor consists of a low power 802.15.4 compliant wireless transceiver, microcontroller and two lead ECG sensor operating with lithium coin cell battery. The limited battery power precludes the possibility of continuous streaming of ECG signal with high sampling rate to the mobile computing platform for analysis. Instead we consider a distributed algorithm that can reliably estimate HRV spectra in the presence of motion artifacts and sensor noise. Our model has two main components: Local computation of potential R-wave locations with associated confidence measures at the heart rate monitor, Postprocessing using Gaussian Process Regression for robust interpolation by probabilistic labeling of True and False Positives at the mobile computing terminal. Potential R-wave locations are computed locally at the heart rate monitor using digital bandpass filtering for extracting QRS complexes, and computing local peaks of the short-time window energy signal. A histogram of the local peak strengths is maintained at the sensor, to assign posterior confidence measures to the extracted positions of R-wave occurrences in the form of a likelihood. Estimated position of R-wave events and associated likelihoods are periodically transmitted wirelessly to the mobile terminal. A Gaussian Process Regression framework provides probabilistic interpolation of the Heart Rate Signal using prior information on the physiology of the heart. In the paper we present experimental results from the distributed wireless HRV system obtained in a laboratory experiment with physical and mental stress events. The results show that the proposed distributed algorithm provides reliable computation of HRV spectra under limited mobility conditions.

(Abstract Control Number: 156)