Machine Vision-based Human Gait Analysis

Abstract

It is important to help elder and injured people to practice the walking process, which is also known as gait rehabilitation or gait training. Quantifying spatiotemporal features of human body motion accounts for a major part of gait analysis refers to measure the body movements in the walking process. The experimentally and clinically data extracted from accurate joint tracking and posture estimate can be applied to calculate the corresponding quantifying spatiotemporal features and construct the gait patterns for helping assess patients with disabilities in walking. However, most of the current gait analysis laboratories are equipped with complicated and expensive equipment such as infrared cameras and floor mounted force platforms and require several cameras working together with various markers locating on corresponding joints of the patient. Therefore, it can be a powerful tool to identify posture-related or movement-related problems in people with injuries or help to improve the performance of athletes by capturing quantified spatiotemporal features through a single monocular RGB camera. In this project, a vision-based gait analysis approach was applied to accurately calculate various gait parameters to aid clinicians in further developing rehabilitation strategy.