Intelligent Liuzijue Action Quality Assessment: A Computer Vision Solution

This study developed a computer vision-based standardized evaluation system for elderly practitioners of the Six-Character Formula exercise. The system utilizes smartphone cameras and an optimized OpenPose framework, combined with a deep convolutional neural network (DCNN) to achieve human pose estimation (HPE), enabling precise capture of multimodal motion features such as limb movements. Through an improved weighted Euclidean distance algorithm, the system generates standardized scores in the 0–1 range, with specialized optimizations for common physiological characteristics among seniors like mild kyphosis and limited joint mobility. By implementing anomaly filtering and time-series interpolation algorithms, the system achieved a 15% improvement in keypoint detection accuracy. In a 12-week clinical trial involving 52 elderly participants, the system successfully tracked significant improvements in standardized exercise scores (median of initial skewed distribution data increased from 0.6957 to mean of final near-normal distribution data of 0.9185, P <0.0167). Compared with traditional manual assessments, this system not only eliminates subjective bias but also enables home-based exercise monitoring through smartphone accessibility. This study innovatively applies computer vision technology to the quantitative evaluation of the Six-Character Formula—a traditional health-preserving practice—providing an objective and accurate motion analysis tool for geriatric rehabilitation. Future research could further integrate 3D pose estimation and respiratory rhythm analysisTechnology, in order to comprehensively enhance the clinical application value of the system.