Gaze Detection System Using RGB-D Sensors for Non-Contact Human Interaction

• Main Authors: Chin-Chen Tsai, 蔡沁宸
• Other Authors: Chyi-Yeu Lin
• Format: Others
• Language: en_US
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/31723835349753241456

碩士 === 國立臺灣科技大學 === 機械工程系 === 104 === In human-computer interaction, gaze orientation is known as an important and promising source of information to demonstrate the attention and focus of users. Within previous research, satisfactory accuracy in head pose and eye location estimation can be achieved mostly in constrained settings. However, currently, real-time gaze orientation based applications are still limited due to low-accuracy or inconvenience of associated head- mounted devices. Also, in the presence of non-frontal faces, eye locators are not adequate to accurately locate the center of the eyes. In this thesis, two novel methods are proposed to improve the existing gaze tracking techniques. The first method uses Kinect v2, one of the latest RGBD devices, to estimate the 3D direction of the head movement and gaze. Different from the previous devices such as Kinect v1 and web-camera, it offers high-accuracy detections and high-resolution images. In the second, a revised pupil search method with optimization is devised to increase the efficiency of searching and so as to significantly minimize the calculation time. Therefore, in this thesis, a hybrid scheme combining the head pose and the eye location information is proposed to obtain the enhanced gaze estimation.