近年來,隨著科技發展,賣場的商品廣告(Advertising)模式已經從過去的報章雜誌與電視,漸漸演變成數位電子看板(Digital Signage),由此可見數位電子看板已被廣泛的運用在許多賣場當中,也改變以往賣場的廣告(Advertising)宣傳方式。 現在許多廠商為提高使廣告效益,開始將顧客的性別與年齡視為推薦的關鍵因素,並針對目標消費群族進行商品廣告的客製化推薦,期望可以達到商品的銷售量。 因此,本研究期望透過人臉的特徵,辨識出顧客的性別與年齡,納入商品廣告的推薦因素,並加入注視點分析結合到賣場數位電子看板的播放系統。希望透過顧客在觀看賣場貨架商品的過程中,分析顧客臉部的性別與年齡作為商品廣告推薦的依據,並分析顧客注視點,來判斷顧客是否正在觀看系統所推薦的廣告。透過這樣的方式,以提升廣告上的效益,改善顧客的互動體驗,達到自動、客製化的商品推薦。 本研究主要開發結合兩種人臉的特徵辨識技術,分別為性別辨識與年齡辨識,透過自動辨識性別及年齡推薦顧客合適的商品廣告。為改善年齡辨識的準確度,本研究提出兩階層的辨識方法用於改善年齡辨識。結果顯示,在人臉的年齡、性別辨識上有相當不錯的辨識率,並且加入了注視點分析,分析顧客眼睛所注視的方位,用來判斷所推薦的商品廣告是否吸引當前顧客。
In recent years, the store's advertising model has been evolved from newspapers and television to digital signage due to the advance of technology and the innovation. We witness digital electronic billboard has been widely used in many stores, which also changed the store’s advertising strategy. Now days, many manufacturers are improving the effectiveness of advertising, and the age and sex of consumer are considered as the key factors in order to target the consumer groups for the commercialization of advertising recommendations, and expected to achieve the good sale. Therefore, this study is expected the face characteristics to identify the customer's gender and age, and put into the recommended factors, and add the gaze estimation together with the store digital signage. It is expected to analyze the gender classification and age estimation and through the process of customer viewing the goods take them as the customer's references. In this study, we analyze the customer's eye moment to determine if the customer is watching the recommended advertise. Therefore, it can enhance the effectiveness of advertising and improve the customer's interactive experience, finally achieve the automation and customization of goods recommendation. It mainly combines the two facial feature identifications, gender classification and age estimation. To improve the accuracy of age estimation, we propose a two-level identification method for improving age estimation. The experiment results show that the age estimation and gender classification have better recognition rate and also was the gaze estimation. For analyzing their eyes gazing position to determine if the recommended advertising attracts the current customers. Finally, we listed some restrictions and improvement methods.