Indoor Occupancy Detection and Estimation using Machine Learning and Measurements from an IoT LoRa-based Monitoring System

In this paper, we present results on the application of machine learning to the detection of human presence and estimation of the number of occupants in our offices using data from an IoT LoRa-based indoor environment monitoring system at Aalborg University, Denmark. We cast the problem as either binary or multi-class classification and apply a two-layer feed forward neural network to the data. The data used for training, validation and testing of the network comprises of environmental data from the IoT sensors and manual recordings of the door and window states. Results show that the classifier is able to correctly determine occupancy of our offices from the IoT sensor measurements with accuracy up to 94.6% and 91.5% for the binary (presence or absence of persons) and multi-class (no person, one person or two or more persons) problems, respectively. Our analysis also shows that occupancy detection with a network trained either in another room or with single environmental parameter is also possible but with less accuracy.