A gaze estimation approach for intelligent viewing in tele-operation

Human like viewing is a major challenge in intelligent remote scenario viewing in tele-robotic applications of mobile as well as fixed frame slaves that are interactively controlled in `man in loop' type control. The control being mainly based on console-views available to the operator, the HMI system must understand what the operator wants to look at. The presented work, attempts at developing a method for guessing operator interest zone in remote workspace area by understanding operator's reaction to `console-view'. The approach establishes a linkage for the remote camera control to subtle operator head movement in reaction to console view. The natural operator reaction is interpreted by our gaze estimator through rapidly extracted image clues.