A vision-based walking posture analysis system without markers

Using a simple method or instrument to realize our health condition not only reduces the cost of medical resources but also achieves the objective of self-health management. This study proposes a vision-based human walking posture analysis system without the need of markers. The system can be applied at home for self-health care or used in health-care institutions. This study analyzes the human walking posture based on side-view and front-view images of a subject. Four features are extracted from the images for walking posture evaluation, including body line, neck line, center of gravity (COG) and gait width. Tilting angles associated with body line and neck line and their periodic variation are adopted to evaluate upper body posture for any abnormality and its correction. The COG and gait width features are used to evaluate the posture condition of lower body parts. Since these two features show an inverse relationship for a normal posture, they can be used to evaluate the stability of a walking posture and see if any posture correction is needed. The experiment results show that this study successfully extracts all the four features mentioned above from the silhouette image of human body. In addition, the proposed system is compared to a traditional marker-based system by performing correlation analysis between their experimental results on the 8 standard decompositions of a full gait cycle, resulting in very high positive correlation. It shows that the proposed walking posture analyzer achieves the performance of a traditional marker-based system with lower cost and higher convenience.