A Video-based Method for Assessment of Hip-Knee-Ankle Coordination during Walking

Coordination of lower limb joints during walking is an important indicator for clinical diagnosis and rehabilitation of various nervous and musculoskeletal system diseases. Despite the significant demand, coordination assessment has not been well explored due to the limited availability of needed measurement equipments and analysis techniques. In this work, we propose a novel method to quantify the coordination of multiple joints taking videos of gait as the input. The locations of joints are derived from the videos using a pose estimation algorithm. The kinematic features such as flexion angles of hip, knee, and ankle are then computed and extracted through the coordinates of the joints. Dynamic time warping (DTW) is performed to overcome the difficulty that the lengths of time series are usually different, and the DTW distance quantifies the coordination. the kinematic parameters of each subject in multiple gait cycles are recorded, and DTW barycenter averaging (DBA) is implemented to synthesize multiple time series into an average sequence, making the method robust. Three-dimensional hip-knee-ankle data are utilized to calculate the DTW distance that quantifies the differences in gait kinematics between individuals. Experiments were conducted with 20 healthy subjects, and the results validated the efficacy of the proposed method. With the advantages of video-based measurement on easy deployment and remote diagnosis, the proposed method promises great potential for clinical assessment of lower limb joints coordination during walking.