Device-free Personalized Fitness Assistant Using WiFi

Abstract

There is a growing trend for people to perform regular workouts in home/office environments because work-at-home people or office workers can barely squeeze in time to go to dedicated exercise places (e.g., gym). To provide personalized fitness assistance in home/office environments, traditional solutions, e.g., hiring personal coaches incur extra cost and are not always available, while new trends requiring wearing smart devices around the clock are cumbersome. In order to overcome these limitations, we develop a device-free fitness assistant system in home/office environments using existing WiFi infrastructure. Our system aims to provide personalized fitness assistance by differentiating individuals, automatically recording fine-grained workout statistics, and assessing workout dynamics. In particular, our system performs individual identification via deep learning techniques on top of workout interpretation. It further assesses the workout by analyzing both short and long-term workout quality, and provides workout reviews for users to improve their daily exercises. Additionally, our system adopts a spectrogram-based workout detection algorithm along with a Cumulative Short Time Energy (CSTE)-based workout segmentation method to ensure its robustness. Extensive experiments involving 20 participants demonstrate that our system can achieve a 93% accuracy on workout recognition and a 97% accuracy for individual identification.