利用人臉辨識探討數位廣告效益的系統雛型

A Face Recognition System Prototype to Evaluate the Effectiveness of Digital Advertisement

10.29943/DNSJ.201006.0283

謝明宏；楊東麟；戴家裕

影像探勘 ； 決策樹 ； 人臉偵測 ； 多媒體數位廣告 ； Image mining ； decision tree ； face detection ； multimedia advertisement

電腦視覺、影像處理與資訊技術研討會

2010（2010 / 06 / 01）

283 - 289

繁體中文

本論文提出一個基於決策樹與人臉辨識技術的人臉目視偵測的方法與系統雛型，可用於多媒體數位廣告效益的探討，例如辨識特定顯示器和時間所播放的廣告有多少人觀看的情況，藉此統計每個廣告對於民眾的感興趣程度。首先系統以視訊辨識技術來辨識人臉、雙眼及嘴，以這些特徵值做為參考資料，再使用我們設計的決策樹模型來判別該人臉的眼睛是否注視螢幕觀賞廣告。我們能夠得到有多少人正在看廣告和觀看時間的長度來做研究，甚至未來加上年齡層、性別等的辨識，將可以應用於瞭解多媒體數位廣告，增進廣告的互動性和便於廣告商的客戶資料蒐集，此系統不僅可以提高廣告效益，也可以偵測當無人觀看時關閉螢幕電源達到節能省電的目的。

In this paper, we propose a face detection method and data collection system prototype. Our system can be used to evaluate the effect of multimedia advertisement. When the screen is showing some advertisements, our system can detect how many people are watching. With the detecting technique, we can gather the statistics for the number of people interested in certain kinds of ads at some period of time. We use the feature recognition technique to identify the people’s face, eyes and mouths. After taking these three features into consideration, we can obtain the information about the number of people who are looking at the screen and how long they have watched. Furthermore, in the future it is possible to recognize the age or gender of the audience. This information can be used to improve multimedia advertisement with personalized interaction or help companies collect the data from their customers. It can not only increase commercial benefits, but also save electricity if the system can turn off the power automatically when no viewer is detected for sometime.