「互動雙足式機器人之設計與實現」是由三位同學合力完成，共分為(A)手勢辨識、(B)雙足式機器人控制以及(C)互動演算法執行三個部分，而本論文乃針對(A)項目做研究。 本論文主要設計一個使用者與機器人互動之手勢辨識系統，使用者以手勢下達指令給主導(Master)與從屬(Slave)兩台stand alone的雙足式機器人，進行獨力搬運、握手、接力搬運、合作搬運之互動的表演或執行前進、後退、左右轉、伏地挺身等基本動作，不同的動、靜態手勢定義不同的指令。在靜態手勢辨識上，辨識之前先將手勢之手臂影像去除，保留手掌之影像資訊。接著以雷達掃描的方式將手勢轉換成角度距離曲線圖，判斷其手指數；再根據手勢手指間角度的差異，辨識相同手指數之不同手勢。在動態手勢辨識上，以先後指尖的位置差異或先後手掌像素的數量差值做為判斷的依據，若其差異大於所設定的閥值，則判定為手左右擺動或移遠移近的動態手勢，本論文總共完成可辨識的九種靜態手勢及四種動態手勢。 The study work “The design and realization of interactive biped robots” was completed by three members. Three members accomplish the following three tasks, respectively, (A) gesture recognition, (B) basic motions’ control of biped robots, and (C) execution of interactive algorithm for two robots. This thesis focuses on the part (A) gesture recognition. The goal of this research is to design a gesture recognition system for human-robot interaction. The operator uses different gestures to command two stand-alone robots “Master” and “Slave” to perform basic motions and four interactive motions. The basic motions are walking forward and backward, turning right and left, shifting right and left, and doing push-up. The interactive motions include 1. Master robot transports the object from the initial position to the destination; 2. Two robots walk close to each other and shake hands; 3. Slave robot passes the object to Master robot, then Master robot transports the object to the destination; and 4. Two robots carry the object together and transport it to the destination. Each dynamic or static gesture is regarded as one command for robots. In static gesture recognition, the first step is to remove the arm of the hand and retain the palm before recognizing. The second step transforms the palm into the curve graph of the angle against to the distance between the center point and each contour point of the palm. By analyzing the curve graph, the number of fingers is determined. The third step is to identify different gestures which have the same number of fingers according to the angle between the recognized fingers. In dynamic gesture recognition, the variation of fingertip’s positions or of the pixel numbers of the palm between two successive images is compared. If the comparison difference is larger than a threshold, the gesture is identified as swinging right and left or moving backward and forward. This research can recognize nine static gestures and four dynamic gestures
