Occupancy Inference Using Pyroelectric Infrared Sensors Through Hidden Markov Models

Occupancy awareness is important for appliance control, ubiquitous computing, and analytics for enterprize real estate. In this paper, a low-cost, battery-powered, wireless occupancy detector equipped with pyroelectric infrared sensor and wireless communication is presented with the aim of providing real-time and on-going occupancy estimation. The methodology adopted to detect occupancy status is based on the hidden Markov models (HMMs). HMMs are trained to statistically estimate the occupancy of a space through an expectation-maximization learning process. The advantage of this system is that the online algorithm reduces annoyance to the users by making it less likely to falsely switch OFF the appliances, and the offline algorithms improve the estimation of the total occupancy time. Simulation and experimental results were obtained to verify the proposed methods along with the limitations of the system.