Detection of Human Walking Posture through a Wearable Camera

This paper proposes to detect leg posture, walking trajectory, and walking speed during walking in the static indoor environment, which are keys to monitor human walk normality. A small wearable depth camera installed on the knee is used as sensor to detect and monitor knee posture including angle, walking distance and trajectory. In this method, points of interest (PoI) are determined and tracked in each frame, and 3D coordinates of those PoI are used to calculated camera angle and moving distance. These are utilized as feature to train and test in a classifier for human walk normality diagnosis.