Human Occupancy From Few Diverse Sensors

This paper addresses the task of estimating the number of human occupants in rooms using data from a small number of sensors, of different kinds, placed in those rooms. In some environments, such as homes, offices, and small stores, the maximum possible number of concurrent human occupants is small (say, 10) and therefore the task may be fruitfully modeled as classification problem. Using robust human-understandable classification methods yields classifiers that provide higher accuracies than those in prior work.