An mmWave Radar Based Real-Time Contactless Fitness Tracker Using Deep CNNs

Abstract

“A healthy mind lives in a healthy body”. To maintain a healthy body, one has to follow a healthy lifestyle. Fitness trackers are excellent companions that help people meet their fitness goals and keep a healthy lifestyle. Thus, fitness tracking is becoming an integral part of the human lifestyle. But, most available fitness trackers nowadays require wearing/mounting on the body, which makes them unattractive for widespread use. To the contrary, mmwave radar sensors can accurately capture micro-motion parameters in a contactless configuration. In this paper, we propose and experimentally demonstrate the use of an mmwave radar sensor for real-time contactless fitness tracking using deep convolutional neural networks (CNNs). In this work, we created a setup to capture the body movements during various exercises by an mmwave radar. The radar data was processed and passed onto a deep CNN for training. Post-training results show that the system is able to classify different exercises in real-time with very high accuracy. The proposed system is computationally inexpensive and has the potential to become an effective alternative to body wearable fitness trackers for indoor fitness activities.