Human movement monitoring and behavior recognition for intelligent sports using customizable and flexible triboelectric nanogenerator

Effective collection, recognition, and analysis of sports information is the key to intelligent sports, which can help athletes to improve their skills and formulate scientific training plans and competition strategies. At present, wearable electronic devices used for movement monitoring still have some limitations, such as high cost and energy consumption, incompatibility of suitable flexibility and personalized spatial structure, dissatisfactory data analysis methods, etc. In this work, a novel three-dimensional-printed thermoplastic polyurethane is introduced as the elastic shell and friction layer, and it endows the proposed customizable and flexible triboelectric nanogenerator (CF-TENG) with personalized spatial structure and robust correlation to external pressure. In practical application, it exhibits highly sensitive responses to the joint-bending motion of the finger, wrist, or elbow. Furthermore, a pressure-sensing insole and smart ski pole based on CF-TENG are manufactured to build a comprehensive sports monitoring system to transmit the athletes' motion information from feet and hands through the plantar pressure distribution and ski pole action. To recognize the movement status, the self-developed automatic peak recognition algorithm (P-Find) and machine learning algorithm (subspace K-Nearest Neighbors) were introduced to accurately distinguish the four typical motion behaviors and three primary sub-techniques of cross-country skiing, with accuracy rates of 98.2% and 100%. This work provides a novel strategy to promote the personalized applications of TENGs in intelligent sports.

Automated summary

Effective collection, recognition, and analysis of sports information is crucial for intelligent sports. This study introduces a novel customizable and flexible triboelectric nanogenerator (CF-TENG) for motion monitoring, which shows highly sensitive responses to joint-bending motions. The proposed system, combined with self-developed algorithms and machine learning, accurately recognizes different motion behaviors and techniques.