A long-term sensory logging device for subject monitoring

Abstract

In this paper we introduce a device for monitoring subjects by logging sensory data over long periods of time. The system consists of a sensory measurement unit, a memory unit and an application for data preprocessing tasks, such as converting sensory measurements to desired units or calculating averages. In order to demonstrate usage, an activity level monitoring system inspired by medical applications is implemented using the device. A rudimentary threshold-based embedded classifier is trained using several different activities, yielding an activity level indicator. 2 subjects are monitored to train the classifier, and the system is then evaluated on new data using those two subjects plus a third not involved in the training process. The results indicate an activity classification of 74% into three levels using 2 simple data thresholds, with a system lifetime of 26 days on 2 AAA batteries.