Testing classifiers for embedded health assessment

Lecture Notes in Computer Science, Proceedings from the 10th Annual International Conference on Smart Homes and Health Telematics
Year: 2012, Vol. 7251, No. , 198-205

Article PDF:

ICOST LNCS - Testing Classifiers 2012.pdf

Abstract:

We present an example of unobtrusive, continuous monitoring in the
home for the purpose of assessing early health changes. Sensors embedded in
the environment capture activity patterns. Changes in the activity patterns are
detected as potential signs of changing health. A simple alert algorithm has
been implemented to generate health alerts to clinicians in a senior housing
facility. Clinicians analyze each alert and provide a rating on the clinical
relevance. These ratings are then used as ground truth in developing classifiers.
Here, we present the methodology and results for two classification approaches
using embedded sensor data and health alert ratings collected on 21 seniors over
nine months. The results show similar performance for the two techniques,
where one approach uses only domain knowledge and the second uses
supervised learning for training.