Collaboration with CNR IEIIT on Indoor Localization and Activity Recognition

Indoor localization and activity recognition from WiFi signals is a field of active research in Communication, Pervasive and Ubiquitous computing domains in recent years. Specifically, fluctuation in signal strength, phase and energy distribution over frequency domains can be used as environmental stimuli to sense location, presence, activities and gestures in the proximity of a wireless receiver such as, for instance, a wifi access point, smartphone or Internet of Things (IoT) devices. For instance, gestures can be observed from Doppler Shifts in a reflected signal, gait of a person moving towards a wireless receiver can be extracted from phase changes in the reflected signals, or even movements as tiny as breathing were visualized exploiting Fresnel effects between a transmitter-receiver pair. However, most of these studies have been conducted in controlled laboratory environments with low interference from the surroundings and with single subjects only. It is an open challenge to improve the robustness of these recognition protocols to work in realistic environments.

With the support of the HIIT open call, Sanaz Kianoush, a PostDoc researcher from CNR IEIIT was invited to visit Helsinki region and in particular Aalto University from 01.02.2018 through 21.02.2018. Sanaz Kianoush’s research interests include Localization in Wireless Sensor Networks, Low-complexity and Energy-efficient localization in Cognitive Radio Networks. Specifically, during her stay, she collaborated with researchers of the Department of Communications and Networking (ELEC, Aalto University) on Multi-antenna systems for device-free activity recognition, localization and counting.

The studies are still ongoing but good progress has been made on the recognition and tracking of gestures from multiple subjects simultaneously. The established research collaboration will be further strengthened in the upcoming months as the studies on the generated data and the design of recognition models is further progressing.