摘要

餵食機器人是一種能夠幫助使用者自主進食的機器人。『進食』是每個人每天三餐必須進行的行為，對於身體障礙者來說，每次進食都必須仰賴他人的幫助，進而造成家人或親友的不便。因此，我們期望能夠透過餵食機器人來取代人力的方式，不僅能減輕家屬的負擔，並且提高身障者的自主性。

以影像為基礎的自動化餵食機器人，是由三個AI-1001伺服馬達(servo motor)以及兩個KRS-2350HV伺服馬達構成的5自由度(5-DOF)機器手臂。藉由CCD攝影機擷取進食時，外在環境的資訊，透過類神經網路系統運算，控制5自由度的機器手臂連結湯匙，規劃、修正手臂移動的位置，把目標食物送至使用者的嘴邊，讓身障者不需要固定進食的位置，僅須憑藉著頭部動作，就能夠達到自主進食之目的。

關鍵字：餵食機器人、嘴唇辨識、頭動滑鼠、倒傳遞演算法、類神經網路

The Development of a Vision-Based Meal-Assistance Robot

ABSTRACT

Meal-assistance robot is a kind of robot to assist the disabled in eating foods by himself. Everyone must eat everyday, but for the disabled, they must depend on other people each time when they eat. It makes caretakers inconvenience. We hope the robot can replace the human to feed the disabled, so that the disabled can be more self-determined and the load of the caretaker can be reduced.
The vision-based meal-assistance robot is the 5-DOF mechanical arm composed of two servo motors KRS-2350HV and three AI Motors AI-1001. The robot gets the environment information as inputs from CCD Camera. It controls the arm with spoon and plans the arm’s path by the way of neural network in order to feed the disabled. The disabled does not use both hand to control the robot, and need not to fix their mouth for eating. They only use head mouse to choose the target of foods which they like, and eat the food freely by themselves.

Keywords: meal-assistance robot, lip detection, head mouse, back-propagation algorithm, neural network