HeartSense

Abstract

Heart rate is one of the most important vital signals for personal health tracking. A number of smartphone-based heart rate estimation systems have been proposed over the years. However, they either depend on special hardware sensors or suffer from the high noise due to the weakness of the heart signals, affecting their accuracy in many practical scenarios. Inspired by medical studies about the heart motion mechanics, we propose the HeartSense heart rate estimation system. Specifically, we show that the gyroscope sensor is the most sensitive sensor for measuring the heart rate. To further counter noise and handle different practical scenarios, we introduce a novel quality metric that allows us to fuse the different gyroscope axes in a probabilistic framework to achieve a robust and accurate estimate. We have implemented and evaluated our system on different Android phones. Results using 836 experiments on different subjects in practical scenarios with a side-by-side comparison with other systems show that HeartSense can achieve 1.03 bpm median absolute error for heart rate estimation. This is better than the state-of-the-art by more than 147% in median error, highlighting HeartSense promise as a ubiquitous system for medical and personal well-being applications.