可偵測觀者注意力之互動展示牆

Abstract

不論是在公共環境或是私人環境中，我們都可以發現許多數位看板。數位看板可以使用LED、LCD、電漿顯示器或投影螢幕來呈現內容。數位看板也可以藉由附加裝置來擴展其互動能力。在我們的作品中，我們使用兩台投影機來投影出高解析度的影像，並在投影螢幕中間架設一台攝影機來攝影螢幕前的影像。我們想知道觀者在觀看投影幕的哪裡，並根據結果與觀者互動。因為視線估測的種種限制，我們採用頭部姿勢估測來估測觀者的注意力所在。此外，我們進行了一些實驗來探討只用頭部姿勢估測來估測注意力會有多少的誤差。 在普普藝術家安迪˙沃荷的作品：Campbell's Soup Cans的啟發下，我們設計了一面互動牆，在牆上有4x11個小屏幕。在牆前有一張具有壓力感應器的椅子，以判定使用者已加入互動中。當使用者坐在椅子上後，壓力感測器便會觸發互動機制，首先相機會取得使用者的高解度影像，並經由臉部偵測裁切出使用者的臉部影像，之後因應不同的互動方式，可以加入不同的技術，例如年齡辨識、表情辨識或臉部相似度測量等。最直接的互動方式便是藉由使用者注意力的位置來引發不同的轉換特效。此互動牆目前已曾在學校、美術館和公共空間中做過展示，也得到了使用者的許多反饋，使我們可以更加改進系統。我們亦提出了一些方法來改善使用者的體驗感。

We can find many digital signs in both public and private environments. Digital signs frequently display content by utilizing technologies such as LCD, LED, plasma displays, or projected images. Digital signage can extend the interactive abilities through the accompanying employment. In our work, we use two projectors to project a high resolution image onto a projection screen, and set a camera in the middle of the screen. We want to know where the viewer is looking at in the image, i.e., the focus of attention of viewer. Because of some limitation of gaze estimation, we use head pose estimation to estimate human attention. Besides, we also conduct some experiments to find out the bias estimating human attention by using head pose only. In applications, we show respect to Andy Warhol, leading figure in Pop Art, and design an interactive wall, which consists of 4x11 arrays of small screens presenting the continuous metamorphosis of “portraits.” When a participant is attracted by the brilliant variation of the portraits and sits on the chair with pressure sensors, his/her face will be captured into the system and activate a series of interactive activities driven by his/her attention. To fulfill the work, several technical components are integrated, including image processing, head pose estimation, pressure detection, and etc. Since interaction driven by human attention is the most intuitive approach, participants can easily engage in the interactive art. After several times of public demonstrations, we have acquired many precious feedbacks from user-experience questionnaires. We also proposed some approaches to improve user experience.