抄録
In recent times, personal mobile devices have become an integral part of daily life. However, since these devices typically demand prolonged durations of focus, they adversely affect a user's blinking rate. Blinking is an essential function of the eye that spreads tears, and thereby, helps avoid eye dryness and soreness. Several blink detection methods have been proposed, however, these methods are unable to effectively monitor involuntary blinks when the camera frame rate is low. To address this issue, this paper proposes a blink detection method for a real-time vision system implemented on the devices that dynamically change the frame rate of the built-in camera according to the usage environment. In the proposed method, the location of eyes is determined and an open-eye template is created by image differencing the initial blinks of the user. Subsequently, the eye is tracked by template matching using the open-eye template in each frame. During eye tracking, the correlation score is analyzed and thresholded with a criterion in order to detect closed eyelids at each frame. Several-frames-of-moving-averaged value of highest correlation score at each frame is used as the threshold value with an offset. In the experimental evaluation, the proposed blink detection method was implemented on a laptop-PC. A total of more than 10,000 true blinks from twenty-three test subjects yielded detection accuracies of 92.2 and 90.1% at 30 and 6 fps frame rate, respectively.
