A Unmarked Knee Joint Angle Calculation Method for Gait Analysis

In response to the existing issues of poor detection accuracy and unstable results in existing unsupervised knee joint angle calculation methods, this paper proposes a knee joint angle calculation method. Firstly, the OpenPose human pose estimation algorithm is used to detect key points in the human body, and the 3D coordinates corresponding to the key points are used to segment the point cloud of the lower limbs of the human body; Secondly, the 2.5D point cloud is shrunk and deformed using the proposed SDMin algorithm to obtain its corresponding bone point cloud; Finally, the direction vectors of the femur and tibia are obtained through line fitting, and the knee joint angle is obtained by calculating the angle between the two direction vectors. Through experimental analysis of the detection accuracy and stability of the algorithm, using OpenPose algorithm, error range of 0.8~4.2° and an average error of 2.31°. However, the SDMin algorithm was used to calculate the knee joint angle, with an error range of 0~1.7° and an average error of 0.76°, and the detection accuracy and stability are significantly improved compared to conventional algorithms.