Extracting silhouette-based characteristics for human gait analysis using one camera
Trong-Nguyen Nguyen, Huu-Hung Huynh, and Jean Meunier
ACM Symposium on Information and Communication Technology (SoICT), Vietnam, 2014
With the strong development of computer vision, health care and in-home monitoring systems are widely applied. Gait analysis is one of the main problems, which needs to be solved in such systems. Most of the recent researches implemented on 3D information of each walking person extracted from stereo cameras or devices with sensors, thus it leads to an increase in the computational cost and price. Therefore, we propose an approach for performing gait analysis using only one normal camera. This paper presents how characteristics are extracted from the walking person's silhouette for gait analysis, in detail, detecting abnormal gaits. Experiments are performed with normal gaits and three different types of anomaly, which consist of hunched back, left-right asymmetry, and sudden motion variation. The obtained results show that there is no case of omission or false detection, and our solution can be integrated into real-time systems.
