Detecting Social Interactions in Indoor Environments with the Red-HuP Algorithm

Detecting social interactions among people represents a challenging task. In this study we evaluate the performance of the ReD-HuP algorithm. We study a real-world and useful experimental dataset and we provide a comparison with some classification methods. Interactions are inferred from co-location of people by exploiting Bluetooth Low Energy (BLE) beacons. Our analysis investigates how the different transmission powers affect the overall performance, we also analyze the results by varying the width of the time window used to analyze BLE beacons. Results obtained with the ReD-HuP algorithm have been compared against two well known and wide adopted machine learning classification methods.