Room occupancy determination with particle filtering of networked pyroelectric infrared (PIR) sensor data

Abstract

Occupancy determination of a large room, such as university campus room, using pyroelectric infrared (PIR) sensors is difficult because PIR sensors are inherently insensitive to humans that are motionless and it is difficult to achieve 100% coverage even with multiple sensors. To overcome these problems, the authors propose using particle filtering of networked multiple PIR sensor data to improve the accuracy of the occupancy state determination. The particle filtering can eliminate incorrect occupancy state transition without incurring large delays. Our method is robust to the incompleteness of the coverage and accuracy. The authors performed an experiment in a campus room (15.3m×6.4m) with 4 PIR sensors to prove the validity of the proposal. The experiment shows that the proposal can improve the occupancy determination accuracy from 29.6% to 98.3%. The system has been continuously used in the university power saving project.