Primary analysis of human's gait and gaze direction using motion sensors

Abstract

On estimating gaze direction of a walking human from images captured by conventional surveillance cameras, his/her head and eyes region in the images are not very helpful; they are usually low resolution, and often hidden by self-occlusion. Way of walking, on the other hand, is more easy to observe because it looks much larger in the images, and is expected to be influenced by gaze direction. Considering this, we analyze the relation between gaze direction and gait by putting multiple motion sensors on his/her body. From the analysis we find arm swing is affected by gaze direction, and this fact supports the possibility of gaze direction estimation from the way of walking.