Mouse Control System Research Based On Eyes Detection Technology

Facial feature detection plays an important role in applications such as face recognition and intelligent human computer interface. In order to help those whose arms are disabled interact with computers without obstacle, 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 location of the operator's eyes in the video image.Sequentially, the system can be divided into three phases, background processing, fiducial position locating and mouse controlling. In the background processing phase, skin tone areas of the background are segmented by skin tone model. The fiducial position locating phase and mouse controlling phase respectively adopts gray-level projection model and binary segmentation model to locate eyes from the partial face rectangle. In the fiducial position locating phase, for every sampling image, after location of the eyes were calculated, two small rectangle images each contains an eye were cut out to calculate the gray-level segmentation value of the eyeballs, then by statistical analysis on all the sampling images, the position of the point in the middle of the eyes with the highest emergence probability is selected as the fiducial position, and the average of the gray-level segmentation values of all the images in which the position of the point in the middle of the eyes locates near the fiducial position is used as the gray-level segmentation value of the eyeball of current operator. In mouse controlling phase, the segmentation value is used to segment the partial face rectangle image, and eyes are located by finding black blocks that are accordant with eyes feature in the binary image after segmentation, thus mouse can be controlled by the face gestures judged by the location relationship of the eyes or between the middle point of the eyes and the fiducial position.Our system is special for the union of two simple eye detection models whose adventures are complementary. Experiments demonstrate the high efficiency of our system in runtime and correct location rate.