- 學位論文
以電腦視覺辨識雙手姿勢為基礎的二維控制介面
A 2D Controlling Interface Based on the Double-Handed Computer Vision Gesture Recognition
摘要
在好萊塢電影“關鍵報告（Minority Report）”的場景中，不但把人類對於未來科技的想像發揮地淋漓盡致，也帶動了一連串相關的研究。若以電腦科技的發展，從早期僅能透過鍵盤操作，經歷滑鼠的發明，到近年來多點觸控的興起，似乎都暗示著未來人機互動將不再侷限於實體操控介面的形式，智慧化的人機互動模式將是未來必然的趨勢。 然而，從相關的文獻當中，不難發現多數的辨識發展仍有其限制與瓶頸存在，普遍運算法需耗費較長的時間。因此，本研究重新省思相關技術，運用連續影像相減及膚色辨識等方法，開發出一套快速的手部辨識與追蹤系統。以揮手「打招呼」的概念定義相關直覺與簡單的手勢指令並結合色彩學習功能以因應不同使用者膚色差異，透過一般視訊攝影機的偵測，即時追蹤使用者在操作時的手勢變化，並以手勢操控物件的縮放與旋轉。藉由本研究的發展，讓使用者能有別於以往的人機互動體驗，以期達到低成本且人工智慧的研究目的。
並列摘要
The Movie “Minority Report” brought out a whole new concept of human computer technology, which has been evoking numerous related researches ever since. The evolution of input devices began with keyboards. There then ensued the invention of “mouse cursors”. Since that decade, multi-touch devices have been all the rage. Thus, more and more HCI technologies have been developed by researchers. According to the related studies, we learnt that there are still technical and conceptual obstacles existing in the recognition process. Hence, we reevaluate the related technical methods, and then we build up a new model of real time two-handed vision-based gesture recognition system based on the existing methods. In addition, the system provides an auto skin tone learning function, and allows users to “rotate” and “zoom” 2D objects on screen.