Research Of Eye-controlled Mouse Based On Visual Tracking Technology

Visual tracking technology has been a hot topic in the field of computer vision. With the advent of high-performance computer and high quality video camera, the visual tracking has been gradually permeating into all areas of our daily life. The combination of visual tracking technology and human-computer interaction will improve the traditional human-computer interaction pattern, bring users a new, natural and efficient interaction experience. The research of eye-controlled mouse based on visual tracking technology is being conducted under this background and has both theoretical and practical significance.Firstly, since the real-time performance is the most important performance indexes for eye-controlled mouse, a fast real-time object detection and tracking algorithm based on Jones-Viola detection framework and Kalman filter is adopted. Specifically, the eye positioning(both coarsely and preci sely) and segmenting is achieved based on anthropometric standards and a template matching algorithm. A test using the Bern face database in five different gazing directions shows that the adopted akgorithm has good performance.Secondly, considering that the "eye blink" is identified as control signal of mouse click action, with the left and right eye representing left and right button respectively, we adopted template matching algorithm to detect the eye blink action. The blink detection accuracy is up to more than 90%, if the appropriate threshold is chosen. Since the movement of the cursor is controlled by feature point’s, based on the geometry principle, a rough estimation method for line of sight is presented, in which the relationship between the feature point and the fixation point of screen is established.Next, based on previous theroy, a piece of eye-controlled mouse application software is developed. The new human-computer interaction system composed of application software and video camera can replace the traditional mouse, achieving the function of the mouse cursor movement and clicking action just by operator’s feature point movement and eyes’ blinking. The test result s approve the good real-time and tracking performance. The developed application software has some application value.Finally, adopting the model-based design philosophy, the cursor tracking system of the eye-controlled mouse is implemented on the DSP development platform based on SIMULINK/OMAP3530. The good tracking performance of the prototype shows its potential for a complete embedded implementation.