The Research on The Extraction of Website Information in Responding to The Gaze

碩士 === 銘傳大學 === 資訊傳播工程學系碩士班 === 105 === Eye tracking applications in recent year are currently the subjects of concern to the public. This paper combined the eye tracking system and the method for locating objects on the webpage in regards to handling the browsing conditions that the general Web use...

Full description

Bibliographic Details
Main Authors: Cheng,Hsiang-Pin, 鄭祥斌
Other Authors: Hwang,Bor-Jiunn
Format: Others
Language:zh-TW
Published: 2016
Online Access:http://ndltd.ncl.edu.tw/handle/91989706353890821831
Description
Summary:碩士 === 銘傳大學 === 資訊傳播工程學系碩士班 === 105 === Eye tracking applications in recent year are currently the subjects of concern to the public. This paper combined the eye tracking system and the method for locating objects on the webpage in regards to handling the browsing conditions that the general Web users confront while navigating the web. The system consists of two parts. First, in order to utilize the eye tracking technology in the multiple fields or broader ranges, this system adopted Kinect Face Tracking Visualization to capture facial features of users as well as their rotation angles of head movements. The system calculated the spatial coordinates and then controlled the PTZ camera with rotation and zooming automatically. This study improved the prior existed visual tracking technique in estimating the visual placements and its accuracy on the estimation had been significantly improved. Second, for locating objects on the webpage, this study used image processing method to find closed blocks on the webpage screen and then correlated to the collected gaze points to identify regions of interest. For the better effects on establishing DOM Tree (Document Object Model Tree), the research study removed unnecessary tags according to the cutting law. The labels that have been kept were referred to as the efficiency of the feature tree. Meanwhile, the collected gaze points further were processed with Gaussian mixture model for the post-processing. The returned visual characteristics and kept feature tree of web content were combined and compared to reflect the truly concerned objects of the webpage, namely as the meaningful objects in this research study. The experimental results showed that the proposed method in this research could be effectively applied to multiple and broader fields due to its better stability for the variable distances and the ability to handle browsing conditions. Keywords:Eye Tracking, Feature Tree, Meaningful Object, Locating Objects on The Webpage, Gaussian Mixture Model