Recgonition of Eye-Gaze Directions for Interactive Man-Machine Interfaces

• Main Authors: Guan-Tong Chen, 陳冠同
• Other Authors: Ching-Hsiang Tseng
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/06849040952329348919

碩士 === 國立臺灣海洋大學 === 電機工程學系 === 97 === The demands for intelligent man-machine interfaces in modern life have made the application of gesture recognition become more and more popular. For example, recognizing human motions in video games (such as the Wii from the Nintendo) and recognizing human faces in security systems are applications of the gesture recognition. Recognition of the eye-gaze direction is an important task in gesture recognition, it can be use to monitor the status of a vehicle driver. By detecting the driver’s drowsy condition and issuing an alert to the driver, lots of potential accidents due to insecure driving can be avoided. The main goal of this thesis is to recognize the eye-gaze direction from video signals of human faces. This can be used in vehicle driver monitoring and interactive video games. We use the color of skin to identify the face area in the image, then locate the eye sockets with an algorithm. Finally, we detect the locations of the pupils within the eye sockets and then use this feature to recognize the eye-gaze direction. In this study, 216 images are extracted from the video signal collected via a web camera. These images contain human faces gazing at 9 different directions (lower right, right, upper right, down, straight, up, lower left, left, upper left). Experimental result indicates that the proposed method can recognize the 9 directions with a recognition rate of 88.42%. Interactive man-machine interfaces which involve eye-gaze directions can benefit from the reseach result.