Kinematic Cycle Consistency Using Wearables and RF Data for Improved Human Skeleton Estimation

Radar-based health monitoring has the potential to enable 24/7 in-home ambient sensing to increase accessibility to care in a privacy-protective, low-cost fashion. This paper focuses on the assessment of mobility and gait for early warning, post-treatment monitoring, and emergency response to health conditions affecting human movement. Current technology has focused on wearables for monitoring personalized health, but wearables are not effective if not worn. In this case radar can provide a backup source of health information, that can be utilized standalone or in conjunction with the information provided by a wearable. This paper explores the estimation of gait parameters and human pose (i.e. skeleton) using RF sensing alone as well as how wearables can be utilized with RF to improve estimation accuracy. In particular, we propose the concept of kinematic cycle consistency using RF and wearable-estimated kinematics to improve skeleton estimation accuracy. Our results are validated using a gold standard Vicon optical tracking system.