ITR: Detecting Activity in Homes with Ubiquitous Sensing to Support Aging in Place

Massachusetts Institute Of Technology, Cambridge MA

Investigators

Abstract

Changes in everyday activities are known to precede declines in health, particularly cognitive conditions and general aging. A sensor system that can detect activities and monitor for changes would enable a new generation of home-based healthcare tools to support preventive healthcare and aging in place. The goal of this project is to develop supervised learning algorithms that detect activities of daily living. This research will test the following assumption: that a large number of low-cost, unobtrusive, and ubiquitously-installed sensors can collect data used by machine-learning algorithms to permit real-time, automatic detection of everyday behavior in the home. A toolkit of tiny, inexpensive sensors that can be easily and ubiquitously deployed into non-laboratory living environments will be designed, tested, and made available to other researchers. Activity-recognition software will be developed to analyze the sensors. output. Evaluation will begin in the laboratory. Unlike prior work, however, sensors and algorithms will be tested in multiple, real households for extended time periods. Our project will evaluate both the performance of statistically-based pattern recognition algorithms and acceptability of the sensors for deployment in living environments of end users for proactive health applications.