Vision-based human state estimation to control an intelligent passive walker

Abstract

The motion of a passive-type intelligent walker is controlled based on visual estimation of the motion state of the user. The controlled walker would be a support for standing up and prevent the user from falling down. The visual motion analysis detects and tracks the human upper body parts and localizes them in 3D using a stereovision approach. Using this data the user's state is estimated to be whether seated, standing up, falling down or walking. The controller activates the servo brakes in sitting, standing and falling situations as a support to insure both comfort and safety of the user. Estimation methods are experimented with a passive-type intelligent walker referred to as “RT Walker”, equipped with servo brakes.