National Conference on Innovative Paradigms in Engineering & Technology (NCIPET-2012) Proceedings published by International Journal of Computer Applications® (IJCA) 24 Face Tracker for Head Position Detection Swati P. Kale Department of Eectronics Engineering Shree Shankarprasad Agnihotri College of Engineering, Wardha, Maharashtra, India. Deepak Dandekar Department of Electronics Engineering Bapurao Deshmukh College of Engineering Wardha, Maharashtra, India. ABSTRACT The driver fatigue detection is one of the most prospective commercial applications of facial expression recognition technology. Current facial features tracking techniques faces three challenges: 1) variety of light conditions and head orientation failure of some or all the facial features, 2) multiple and non rigid object tracking, and 3) facial feature occlusion. In this paper, we propose a new approach. First, the single camera (webcam) is used to detect face under various lighting conditions. The detected face is used to track facial features by using color model. Because color processing is very fast that mean time requirement is less. And from tracked facial features we predict the head motions in up-down and left-right direction. Furthermore, face movement are assumed to be smooth so that a facial features can be tracked with three point algorithm. Simultaneous use of YCbCr color mode, three point algorithms and the Geometric model greatly increases the prediction accuracy for each feature position. The experimental results shows validity of our approach to a real life facial tracking under various light condition, head orientations and facial expression. General Terms YCbcr color model, three point algorithm Keywords Face detection, Facial feature tracking, color model, geometric model. 1. INTRODUCTION The trucking industry, highway safety advocates, and transportation researchers have all identified driver drowsiness as a high priority commercial vehicle safety issue. Drowsiness affects mental alertness, decreasing an individual‟s ability to operate a vehicle safely and increasing the risk of human error. Successfully addressing the issue of driver drowsiness is a multi-faced challenge. Operational factors such as work schedules, duty times, rest periods, recovery opportunities, and response to customer needs can vary widely. In addition, the interaction of the principal physiological factors that underlie the formation of sleepiness, namely the homeostatic drive for sleep is complex. Thus researches on detecting driver‟s state effectively and preventing accidents are of active meaning. There are two methods to address fatigue of driver. One is to detect driver‟s behavior from vehicle side. It is nothing but the monitoring the steering, velocity of vehicle, latitudinal acceleration of vehicle etc. The other is to detect physiological signal and physical side of the driver. Physiological signal includes heart rate, body temperature brain wave etc. And physical analysis includes tracking of eye, yawing status, head orientation etc. 2. PROBLEM DEFINATION From above method addressing of a driver from vehicle side and detection from physiological signal is expensive method as compared to detection of driver from physical side. And also complexity of the system is increases. Therefore, we address driver behavior from physical side by using a single camera. In-vehicle technological i.e. addressing a driver from physical side approaches has effective tools to address fatigue. Sleepy drivers exhibit certain observable behaviors, including eye gaze, eyelid movement, pupil movement, head movement, and facial expression. The focus of this paper is on the last category of alertness monitoring technologies. This system continuously runs a face finding algorithm which ensures automatic subject calibration and re-acquisition. Then automatically generates a face model that takes into account unique facial features. This process takes less than a four second. Once this model is created, the system begins 3D head pose tracking and attention. Head pose is tracked to +/- 90 degrees of rotation. The system then finds status of the driver, monitoring the frequency head position and alarm system. All data is then output for prediction purposes analysis. 3. HEAD TRACKING SYSTEM 3.1 Problem Definition 1) Localization of face by using camera which is placed in front of the driver. 2) Detection of face by using color model because color processing is faster. 3) Tracking of facial features by using Hough Transform. 4) Develop Geometric face model by using tracked facial features. 5) Estimation of motion by using three point algorithms. 6) Addressing driver‟s status from output data analysis. 3.2 System Design First video stream taken as an input. By applying color model driver‟s face is detected [1]. Then skin model images are binarized i.e. converted into 0 and 1 form. Afterwards the morphological processing and region growing operation are done on binary image to obtain coordinate data of every region. In this way face region picked out from driver‟s face. Then reduce face detection region to improve speed of the active facial tracking system. The implemented system then generates a face model that takes into account unique facial features. This process takes less than a second. Once face model is created, the implemented system begins head pose