Proceedings of AMBIENT 2015, The Fifth International Conference on Ambient Computing, Applications, Services and Technologies
Numerous applications of ambulant medical care, house automation and security use binary sensors such as passive infrared motion sensors or light barriers to monitor activity in the house. Multi-target tracking algorithms allow for at least a partial separation of activity in data from such sensors from multiple persons. While many tracking algorithms demonstrate good performance across various sensing modalities and sensor setups, little research has been done to determine the impact of placement and varying density of sensors for tracking performance. This paper presents the results of an evaluation of a Bayesian multi-hypothesis multi-target tracking algorithm on data of two residents monitored by a network of binary sensors. We evaluate the algorithm on data from sensors of varying quantity and placement. We show that our approach outperforms other approaches in low-resolution setups. While tracking performance naturally decreases with the number of sensors, it also strongly varies by sensor positioning.