Research on Measurement System of Upper Limb by Using Single Monocular Web Camera and Inference AI

This study introduces “UpperLimbs,” an advanced upper limb motion measurement system that utilizes a monocular web camera and inference AI to simplify the traditional elaborate setups required for three-dimensional motion measurement. Using the MediaPipe AI framework, this system is capable of identifying joint positions and depth information from a single camera's video feed. The efficacy of UpperLimbs was validated against a traditional marker-based system, demonstrating comparable accuracy and potential for significant improvements in rehabilitation and prosthetic training. Further enhancements were explored through various image processing techniques, including image sharpening, noise reductions. The implementation of low-pass filters in time domain and calculated offsets further optimized the system's accuracy, making UpperLimbs a promising tool for accessible and efficient upper limb motion analysis, particularly for individuals using prosthetic hand.