本論文提出一個雙手自然使用者介面,本系統可以追蹤使用者雙手的手掌及手指,並且根據手掌及手指的資訊來辨識使用者雙手的狀態,並進行雙手手勢的辨識,以供使用者進行人機介面的操作。 本論文提出的雙手自然使用者介面共分單手操作以及雙手操作,共兩種操作模式;單手操作的模式時,當使用者的手勢為五根手指併攏時,系統則進行手勢移動方位的判斷,本系統共可判斷八個方位之手勢移動方向,此外,若手勢不為五根手指併攏使時,則為判斷手指是否進行點擊動作,使用者可以藉由點擊圖形化界面上的按鈕來進行想要的系統操作;當系統偵測到為雙手操作時 ,使用者可以使用特定的手勢來進入縮放模式,一旦手勢成立則可以利用改變兩手之間的距離來進行縮放的操作。 並且我們邀請多位使用者協助測試系統之準確率,其移動方向偵測準確率為97.22%,且手指點擊偵測之準確率為93.30%,並且雙手縮放偵測準確率為91.25%。最後我們也將所提出的系統進行硬體架構設計,以數位積體電路設計流程實現為數位晶片,讀取到視訊攝影機送出的影像後,晶片將偵測出手掌座標、手指座標以及手指長度資訊,做為手勢辨識與手指點擊的判斷依據。
In this thesis, we proposed a “Two-hands Natural User Interface”. Our system could track two hands and fingers of user. Furthermore, the system recognized the state of user’s hands via information of palms and fingers. Our system provided a intuitional Human-computer interaction(HCI) for users. Our two-hands natural user interface included both single and two-hands operation mode. In the single hand operation mode, the system would detect the eight kinds of hand moving direction while user’s gesture is the “united finger”. But that the finger isn’t united, the system would detect the finger clicking of user. In finger clicking state, user could select the operation of virtual button via clicking that button. In the two hands mode, the system would detect the zoom in and zoom out while user’s gestures of two hands are the specific gesture. The system detected the specific gesture via the fingers information of finger detection. User could operate the zoom via the gesture of two hands. We invited several users to test the accuracy of system. The accuracy of the moving direction detection was 97.22 %., and the accuracy of finger clicking detection was 93.30 %. The accuracy of two-hands zoom detection was 91.25%. Finally, we implemented our system into a digital chip via hardware architecture and cell-based design. After video camera sent data of image, our chip would detect the palms coordinates, fingers coordinates and fingers lengths for action recognition.