ABSTRACT
ABSTRACT: Eye gaze is an ideal input channel to construct convenient and natural Human-Machine Interfaces (HMI). However, the Midas Touch problem must be taken into consideration under the gaze-based HMI systems. In this paper, We incorporated the visionbased object recognition method into the traditional gaze-based HMI system to avoid the error commands caused by unintentional gazes. A double selection was designed to produce different commands. Dwell-time-based approach could select a possible input gaze and extract the ROI (the region of interest). Object recognition based on shape matching could avoid errors caused by noninput gazes and recognize the gazed object. A mobile eye tracker with a scene camera was used to detect the eye gaze as well as the objects needed to control. The system proposed in this paper was applied to control a real air conditioning. Four functions of switch, mode, heating and cooling could be controlled by gazing four different icons. The results showed that the proposed eye control system could provide more control degree and reliable accuracy with the incorporation of the vision-based object recognition method.
