手勢辨認在電動輪椅控制的應用

Abstract

本論文發展出一套手勢辨認與控制系統，並針對遠端電動輪椅控制設計了合乎自然操作方式的操作手勢。操作者可依近似於推行普通輪椅的操作方式，操控遠端之電動輪椅。 手勢辨認技術以粒子濾波器追蹤雙手手掌的位置，系統再依據雙手手掌的移動位置與手掌開合的狀態以辨認手勢，並產生出相對應的命令。該粒子濾波器採用了適當的特徵，以容忍在追蹤過程中手掌開合時產生的外型變化。並且本系統追蹤雙手時，增加特殊的處理方式，使得雙手間不易產生交互影響；同時也防止了環境光源改變時對可能產生追蹤錯誤的情形。 實驗結果顯示本論文提出之方式，確能達到操控遠端電動輪椅之需求。在執行效率、辨認命令的正確度上都有極佳的表現，同時亦能抵抗各種外界干擾的情形。

A Gesture Recognition and control system is developed in this paper. The system is applied to control remote electric wheelchairs with natural gestures. The way users control remote electric wheelchairs is similar with the way users control general wheelchairs. Gesture Recondition dependents on motion of hands and open-closed hands detection, then the system generates command according to gestures. The motion of hands bases on particle filter tracking which selects proper features to tolerate shape change. While hands tracking, the system has specific handling methods to avoid hands interference and overcome tracking fail due to environment lighting change. The experiment results show that the system is suitable to control remote electric wheelchairs. It also has good performance in efficiency, accuracy and the ability to resist interferences in different events.