A Model-Based RF Hand Motion Detection System for Shadowing Scenarios

With the explosive growth of mobile computing, new modes of human-computer interaction (HCI) are emerging and becoming feasible. Compared to vision-based systems that require lighting, radio frequency (RF)-based hand motion detection systems are becoming more popular in HCI applications. In real RF hand motion detection scenarios, the line-of-sight between the transmitter (Tx) and receiver (Rx) is usually blocked. Hence, shadowing significantly affects the detection accuracy. To design better RF hand motion detection systems, we propose a simple diffraction and interference model (DIM) to interpret the received signal strength (RSS) variation caused by hand motions in the shadowing scenario. Based on theories of knife-edge diffraction and mutual radio interference, DIM provides a simple theoretical foundation for analyzing the RSS variation with hand size, signal frequency, and Tx-Rx distance. Furthermore, a model-based RF hand motion detection system benefiting from DIM is presented. Unlike existing systems that require a large number of motion features to train a motion classifier, the model-based system achieves training-free motion classification, which has potential for hand motion detection on a real-time basis. Empirical data collected from a vector network analyzer validate our system as well as demonstrate a simple diffraction model can help hand motion detection processing for commonly growing HCI applications.