Monitoring and interpreting human movement patterns using a triaxial accelerometer

Abstract

This thesis addresses the hypothesis that a single, waist-mounted triaxial accelerometer (TA) can be used to monitor human movement patterns in unsupervised, freeliving subjects over extended periods, and that it can be used to quantitatively measure parameters that can provide clinical insight into the health status of the subject. A rigorous theoretical and experimental understanding of the signals obtained from a TA is developed. The effect of the placement of the TA device on the waist is explored and a model relating device position to TA signal is developed for a range of postures and activities. A classification framework for movement identification using the signals from a waist-mounted TA is presented. This framework is based on a hierarchical binary processing tree and is designed for real time use. An implementation of this framework for monitoring housebound patients is presented. Algorithms for detecting falls, distinguishing between activity and rest, classifiying transitions between different postural orientations, and for identifying periods of standing, sitting, lying and walking are developed. In evaluation studies performed in controlled laboratory conditions, every algorithm performed with better than 90% accuracy. Once movements are identified, movement-specific parameters sensitive to changes in functional status are extracted from the signal. A two stage methodology for employing the accelerometry system in monitoring free-living subjects is introduced. The first stage involved monitoring specific movements through a directed routine. The second stage involved monitoring of free movement. Signals obtained from the directed routine are used to extract clinically relevant, movement-specific parameters. Signals obtained from the period of free movement are monitored for falls and other abnormal events. General parameters of movement, including energy expenditure, are also measured. The system was evaluated in a series of field studies in laboratory and home environments, in supervised and unsupervised settings, using cohorts of healthy subjects. A pilot trial was conducted in which six healthy elderly subjects wore the TA device for a period of up to three months. The technical performance and useability of the system were evaluated. Clinically significant parameters were measured and the effects of age and health status on the measured parameters were evaluated.