Cluster Based Gaze Estimation and Data Visualization Supporting Diverse Environments
Chiao-Wen Kao, Huihui Chen, Shyi-Huey Wu, Bor‐Jiunn Hwang, Kuo‐Chin Fan · 2017
This paper proposes a method to explore the navigation of audience by estimating gaze points and labeling them to the objects of video or web content. The cluster based gaze estimation and statistics based labeling method are proposed to support the multiple user environments as well as improve the accuracy and usability. In addition, the number of feature clusters is accorded as the amount of areas of target screen. Additionally, statistics based labeling method is proposed to mapping the gaze points to target object by computing the probability. Moreover, this method can provide a good manner to estimate the attentive objects. The visualization of attentive blocks and the amount of probabilities are presented to illustrate the navigation of audience as attentive object. Therefore, this visualization method can exhibit the navigation behavior of audience more real, besides overcomes the problem of difficult labeling objects. The experimental results demonstrate the proposed method provide more robustness for long range as well as diversity environments.