SBIR Phase I: Artificial Intelligence (AI)-Powered, Wearable Technology to Monitor and Assess Strength Training Activities

The broader impact /commercial potential of this Small Business Innovation Research (SBIR) Phase I project is in developing improved assessment tools and as a result, improved knowledge about muscle health and strength training. The Center for Disease Control (CDC) data shows that 76 million people in the US do strength training consistently and meet exercise requirement standards.

The social and health benefits of strength training like muscular growth, fat reduction, better body balance, and improved mental health have been well characterized in numerous peer reviewed studies. There are many wearable devices on the market that can monitor cardiovascular health via heart and respiratory rates, but muscle health data is missing, because monitoring strength training is much more difficult.

This project will develop a low-profile, artificial intelligence (AI)-driven wearable system composed of wrist and torso sensors. The proposed system will automatically recognize a training program, learn about the user goals and training experience, and provide step by step guidance, individualized to the user's abilities and activities. The solution will work in any environment (e.g., home, gym, or outdoor location). The system will be able to initiate, maintain, and improve strength training health regimens.

This Small Business Innovation Research Phase I project will combine AI and signal processing techniques to enable inertial measurement unit (IMU)-based sensors to robustly detect strength training movements, accurately count repetitions, and provide performance metrics like time under tension, exerciser pace, amount of the total work during an exercise, power generated, range of motion, etc. Towards this end, an optimized deep-learning model will be built that will detect 15 strength training exercises with 99% accuracy and will miss no more than one out of hundred repetitions.

Another key objective is the creation of a neural network structure to reduce IMU drifts via feature aggregation of the knowledge of an exercise type being performed and human kinematics. The anticipated technical result is performance assessment with a root-mean-square deviation of less than 0.02 m for trajectory and less than 0.025 m/s for velocity.

Finally, anticipated results include documentation of user needs aligned with implemented and future product features.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.