The Study Of Mouse Control System Based On Facial Feature Detection

"Information accessibility"is defined as any person in any environment can enjoy the convenience brought by the information technology, to improve work efficiency and quality of life. Benefited people extend to the elderly and other healthy people. In order to make the upper-limb disabled share information resources, we develop a mouse control system in which video camera is used as input device instead of traditional manual mouse, and the mouse is controlled by locations of the operator's eyes and lips in the video frames detected by facial feature detection.We introduce the system sequentially which can be mainly divided into three phases, background processing, facial feature locating and mouse controlling. In the background processing phase, skin tone areas of the background are segmented by skin tone model to be differenced later. In the facial feature locating phase, for every front face sampling image, after location of the eyes and lips were calculated by gray-level projection model, small rectangle images contain eyes and lips were cut out to calculate the gray-level segmentation threshold, then by statistical analysis on all the sampling images, the positions of eyes and lips with the highest probability are selected as facial feature location information, and the average of gray-level segmentation thresholds of all the images around the location is used as the gray-level segmentation threshold of the eyeball of current operator. In the mouse controlling phase, because colors of eyeballs and lips are different with skin tone area, the center positions of eyes and lips can be detected by binary segmentation model and lip color model, and connecting the three center positions we get a facial feature triangle. Compared the real-time triangle parameters with the pre-front face triangle parameters, the face actions including turning and inclining can be estimated, which are used to control the mouse later.Rough face recognition and image normalization are involved in the preprocessing of the system, which improve the process of detection and judgment. Rely on the detection of facial feature triangle, the system achieved the deep rotation of the face, which make the operator more comfortable. Experiments demonstrate the high speed and efficiency of our system which is also robust.