Development of Yawning Detection Algorithm for Normal Lighting Condition and IR Condition

Abstract

Drowsiness monitoring system has been widely used in this current technology to monitor the driver’s state while driving. This paper presents a drowsiness detection method through the activity of yawning for both normal lighting condition and Infrared (IR) condition. Development of the algorithm consists of several steps. Initially, the detection of face and mouth implementing the Viola-Jones algorithm takes place. For IR condition, the mouth is detected by applying the geometrical measurements of a face. After the detection process is done, the tracking process for both face and mouth takes place utilizing the Kanade-Lucas-Tomasi (KLT) algorithm which is basically a point tracking algorithm. Based on the tracked mouth, the region of interest (ROI) is selected which is to be used as an input image in the image processing step in order to get a clearer image of the mouth. From the finalized mouth image in the preprocessing step, the properties of the image are further used in the yawning detection step. In the indication of yawning, the height of the mouth opening reading score is observed. The performance of the proposed method is tested on 5 subjects and achieved an overall accuracy of 98.89% for normal lighting condition and 95.29% for IR condition.