5G-Based Passive Radar Sensing for Human Activity Recognition Using Deep Learning

Wireless sensing and human activity recognition is an important research direction for next-generation wire-less communications systems as they evolve towards Integrated Sensing and Communications (ISAC) using a single waveform. Although gesture recognition is a well-studied field in WiFi systems, relatively few works have exploited the use of the Fifth Generation (5G) New Radio (NR) waveform for sensing and activity recognition. This paper explores the utilization of 5G NR as an illuminator of opportunity to perform passive radar detection and activity recognition. It proposes leveraging the synchronization signal block (SSB) signal of 5G, which is emitted periodically, for processing in passive coherent location (PCL) systems. We develop a signal processing pipeline involving processing of the SSBs that are transmitted at regular intervals to extract the Channel State Information (CSI) between the transmitter and the receiver. The CSI data which provides the characteristics of the radio channel is fed into a Deep Learning (DL) model to discern specific gestures. Experimental results with an over-the-air 5G signal collection campaign show excellent classification results across five categories of activities.