ANALYSIS OF HUMAN WALKING POSTURE USING A WEARABLE CAMERA

. In recent years, human posture has received widespread attention in the field of health promotion, and walking is one of the most popular exercises for keeping daily health. With respect to the walking posture of a person especially of the elderly, small changes in the posture have a great impact on their spine and lower limbs, resulting in changes in balance function. According to this fact, it is necessary to check human walking posture periodically. For this purpose, a simple, automatic way of examining a human walking posture is expected. This paper proposes a method of analyzing human walking motion: Two kinds of indicators about human walking posture are proposed which estimate posture changes from a camera worn by a person through changing scenery observed from the camera. In this method, AKAZE is applied to images to detect feature points and find their correspondences. A 5-point algorithm is used to estimate epipolar geometric constraints and provides an essential matrix for relative camera motion. Changes in the relative motion of the camera are used to analyze the changes of human walking posture. In the experiment, the characteristics of four types of walking postures (one is normal and the others are biased) were analyzed using the human posture indicators, and the reliability of the proposed method was verified.