Human complex joint motion recognition based on plastic optical fiber wearable system

Human Activity recognition (HAR) has significant research value in fields such as movement behavior analysis and health monitoring. This paper proposes a wearable system based onD-type plastic optical fiber (D-POF) sensors. The combination of transfer learning (TL) and the Gramian Angular Field (GAF) -ConvNeXt algorithm has achieved high-precision recognition of complex joint movements in the human body. The developedD-POF sensor features extremely high sensitivity, excellent repeatability and resistance to environmental interference. It is embedded in shoulder, elbow and knee guards to collect motion signals. Subsequently, through the GF-ConvNext network structure, TL is introduced to classify and identify the motion data. The experimental results show that the recognition accuracy rate of the scheme based on theD-POF wearable system and recognition algorithm for four types of complex joint movements reaches 98.75 %. This study verified the high precision and reliability of wearable systems based on plastic optical fibers in daily activity monitoring, providing a new technical solution for the field of intelligent recognition.