OpenPose-based Gait Analysis System For Parkinson’s Disease Patients From Arm Swing Data

The number of Parkinson’s Disease (PD) patients is expected to increase with the rapidly aging society. In this light, this study develops a PD diagnosis system by using the bilateral differences between left and right arm swings as estimated using OpenPose. This system is expected to easily and automatically diagnose the early stage of PD. We used videos of patients who did and did not take anti-PD medicine as well as individuals not diagnosed with PD. By comparing and analyzing the arm angles from these videos, we could distinguish arm swings that reflected PD symptoms from normal arm swings. This study had the following drawbacks: the videos featured only elderly males and were limited in number. To overcome these drawbacks, more diverse videos could be analyzed in the future.