Why Your Heart Was Beating: Poster

Abstract

The use of wireless electrocardiograms, wearable as well as implants, enable long-term and unobtrusive monitoring of patients in their everyday living and working environments. If enriched by environmental contexts, these devices can be vital for early detection of cardiovascular diseases. Often cardiologists encourage patients to keep medical journals in order to contextualise the measurements of electrocardiograms. Experiences show, however, journal entries can be inconsistent or incomplete. In this paper we associate the measurements of a wireless electrocardiogram with the measurements of inertial sensors in order to reason about the activities of a person. We put together the raw measurements and their wavelet transform in a three-way tensor and apply tensor decomposition to uncover hidden features which can be vital for detecting the underlying activities. We model and reason about six everyday activities, namely, cycling, climbing up and down a staircase, jumping, push-ups, running, and skipping.