Face Registration Using Wearable Active Vision Systems for Augmented Memory

This paper describes a wearable active vision system called VizWear-Active in which an active camera is used to obtain more information about the wearer and his or her environment for a wearable vision system. We have constructed the prototype system based on VizWear-Active and implemented two reflex actions: gaze direction stabilization and active tracking. For the gaze direction stabilization, the direction of the camera-head is controlled using an inertial sensor, which reduces the influence of the wearer’s motion on the input images. For the active tracking, the system tracks a person by controlling the direction of the camera-head to observe the person even if the attention of the wearer is focused else where. Our prototype system performs for these reflex actions in real time. Autonomous face registration was implemented on our prototype system for visual augmented memory applications. Facial images can be used to retrieve visual memory cues related to a human subject if various facial expressions are registered in face dictionary. Our system automatically registers the facial images in the its dictionary and uses them to retrieve visual memory cues when the system finds a particular person. We confirmed the basic functions of autonomous face registration in experiments.