A Thermal Data Simulation Tool for the Testing of Novel Approaches to Activity Recognition

Abstract

Researchers require access to sensor datasets for the development of novel data driven activity recognition approaches. Access to such datasets is limited due to issues including sensor cost, availability and deployment time. The use of simulated environments may facilitate rapid generation of comprehensive datasets with minimal cost. This paper introduces an approach to the simulation of thermal sensor data for activity detection and recognition. The approach utilizes multi-touch interfaces to facilitate intuitive recordings of a range of scenarios and supports deployment to a range of mobile platforms. Functional testing has considered the ability of the approach to record activities including: normal room navigation, falling, hypothermia and multiple occupant navigation.