許多身心障礙者於日常生活中有諸多不便，有鑑於此，本研究所屬之輔具團隊，已發展出一套輔助性電腦輸入裝置，提供身心障礙者操作嘴控開關，透過摩斯密碼來控制鍵盤及滑鼠，並也發展出一套電子輔具系統，其系統包含家電控制、環境控制、安全警告等。然而在使用這些系統前仍需他人協助放置嘴控開關，因此本團隊陸續提出多套自動定位系統，來輔助使用者，但這些系統仍有些小缺陷。譬如機械手臂體積過大收納不易，機械手臂過於輕巧，導致奶嘴放置時，安全性不佳或是垂放奶嘴時，放線過多等問題，然而本研究則提出較為合適的機械手臂大小，以及稍精確性的感測定位與影像追蹤，也改善了放線過多之問題，由情境資訊來得知此系統的狀態，當狀態有問題時，工程人員可由雲端資料庫得知，則可對此系統進行維護，以保障使用者與此系統之安全性。
本系統由電腦、機械手臂、視訊鏡頭、慣性感測元件所組成，主要分為三大部分，一為機械手臂運動，透過正反向運動學中的幾何轉換進行座標轉換，來控制機械手臂的移動。二為影像處理，藉由視訊鏡頭將影像擷取，經由PC來進行影像處理運算，先進行人臉辨識標記出人臉特徵部分：眼睛、嘴巴，透過連續影像相減法將動態物件標記出，而在利用粒子濾波於影像中明確追蹤奶嘴動態方式，來進行放置奶嘴的動作。三為情境感知，利用慣性感測元件來感測手臂移動時的座標變化與輔助感測定位，作為情境資訊，將此資訊存入資料庫讓工程人員分析，判斷此系統是否有問題，以達到情境感知的目的。
本系統於放置與收取奶嘴過程，影像處理部分可明確的標記奶嘴位置，利用粒子濾波再次確認此位置，在機械手臂移動的過程則由慣性感測元件來感測座標，其感測的誤差範圍在±3cm，然而當有無法辨識時，工程人員則可透過資料庫數據可知來前往維護此系統，在驗證系統功能與臨床測試過程中，以模擬人臉圖像進行測試，在放置與收取奶嘴部分，平均成功率可達90%，而在脊髓損傷個案的臨床測試於場地不同使得成功率降低，調整後成功率可達80%，可見本系統在功能性與實用性皆已具備，但仍然有改善的空間。
本系統於未來可加強於輔助感測定位的精確性與情境感知的部分，縮小感測的誤差範圍，讓此系統達到更準確的定位，情境感知部分則可增加不同感測功能，讓機械手臂可提供更多元的服務。

People with disabilities have inconveniences in their daily lives. As the result, our assistive technology team had developed auxiliary computer-input device which can enable them to control the mouth switch. Afterwards, they can control the switch to control the keyboard and mouse with the Morse code. In addition, an electric aid, the system which includes home appliance control, environmental control, and safety warnings. However, before the disabled use the computer, the assistive computer input device still has to be placed into the mouth of the user by family or other. Therefore, our team had proposed many automatic position systems to help users, but these systems still have some defects. For instances, the robotic arm is to huge to store in the house, probably the structure of the robotic arm is not stable which will cause the problem of security and position. We propose a new robotic arm system which has appropriate size for user, moreover, come out with sensing position and image tracking to help solving position problem. We use Context-aware information to keep tracing the status of the system. As the result, through cloud database engineers can detect the problem and recover the system on time to ensure the security of the system.
The robotic arm system consists of computer, robot arm, video camera and inertial measurement unit. Each of them extends to one topic. First, a robotic arm movement includes geometric concept of forward kinematics and inverse kinematics to transform the coordinate axis and control robotic arm movement. The second part is image processing. By the images which were captured by video camera, computer can do image processing operations. Including face recognition to know facial features (e.g., eye, mouth.., etc.) and using temporal differencing with particle filter to track dynamic object, and put down object. Last, using IMU(Inertial Measurement Unit) to get moving information of robotic arm and establish the context awareness system which can help correcting the movement of the robotic arm. That is, this context awareness system can enable engineer to maintain the robotic arm immediately.
During robotic arm pick up nipple or put down nipple, image processing can clearly mark location of nipple, and use particle filter confirm this location. The IMU can know the motion of the robotic arm. The IMU in motion of tolerance is ±3cm. However, if the image can’t process, engineer can see the cloud database to know whether robotic arm has any error. The accuracy of the system is average 90% when it tested clinically for homemade face image. Furthermore the accuracy of test for the subject with severe spinal cord injury is also 80%. Obviously, the system is complete in functionality and practicability although it still has some defects.
The system can be improved by strengthen sensing location and context-aware in the future. By reducing tolerance of IMU, this system can achieve a more accurate position. The context-aware information can increase more different sensing function, so that the robotic arm can provide more services.

[1]C. M. WU and C. H. LUO, “Morse Code Recognition System With Fuzzy Algorithm For Disabled Persons”, J. Med. Eng. & Tech., Vol. 26, Num. 5, pp. 202-207, 2002.
[2]C. H. Liang, A Morse Code Based Electronic Aid for Daily Living and Its Application to an Assistive Multi-Agent System, Cheng Kung University,2010.
[3]陸偉杰，”基於ZigBee無線網路之輔助環境監測與控制系統”，國立成功大學碩士論文，中華民國100年2月。
[4]陳珮芬，”ZigBee無線數位家電之摩斯碼嵌入式系統設計”，國立成功大學碩士論文，中華民國100年7月。
[5]羅錦興、梁志賢、陸偉杰(2009)，輔助性人因導向智慧型代理人系統，台南：國立成功大學。
[6]李振賢，”人眼定位之自動奶嘴開關載運系統”，國立成功大學電機工程學系碩士論文，中華民國98年7月。
[7]陳正賢，“具臉部辨識機制的機器手臂於奶嘴開關運送之運用”，國立成功大學電機工程學系碩士論文，中華民國99年7月。
[8]王健宇，”可變形樣板於即時人臉追蹤系統之應用”，國立中山大學機械與機電工程學系碩士論文，中華民國96年7月。
[9]陳宣銘，”Non-Photo-Realistic Illustrations with Artistic Style”，國立中央資訊工程學系碩士論文，中華民國92年12月。
[10]曾郁展，”利用人臉五官為特徵之人臉辨識系統”，國立中山大學電機工程學系碩士論文，中華民國94年6月。
[11]International Federation of Robotics(IFR), http://www.ifr.org/
[12]B. Schilit and M. Theimer. “Disseminating active map information to mobile hosts”. IEEE Network, Vol. 8, No. 5, p. 22-32. (1994)
[13]B. Schilit and N. Adams and R. Want. ”Context-Aware Computing Applications”.IEEE Workshop on Mobile Computing Systems and Applications. Santa Cruz, CA, US, IEEE Computer Society Press.(1994)
[14]A. Schmidt, M. Beigl and H.-W. Gellersen, “There is more to Context than Location“, Computers & Graphics, Elsevier, Volume 23, 1999.
[15]A. K. Dey ,”Understanding and Using Context”, Journal Personal and Ubiquitous Computing,Vol.5,No.1,pp4-7,Feb 2001.
[16]蔣國祥，"以FPGA為基礎實現且應用於機械手臂之影像辨識系統"，國立成功大學電機工程學系碩士論文，中華民國102年1月。
[17]P. Viola and M. Jones, "Robust Real-time Object Detection", International Journal of Computer Vision, pp. 1-25, 2001.
[18]P. Viola and M. Jones, “Rapid Object Detection using a Boosted Cascade of Simple Features,” IEEE CVPR, 2001.
[19]K. E. Clothier and Y. Shang, "A Geometric Approach for Robotic ArmKinematics with Hardware Design, Electrical Design, and Implementation", Journal of Robotics, Volume 2010,10 pages
[20]R. Szeliski,” Computer Vision Algorithms and Applications”, Springer, March 12, 2010
[21]F. Gustafsson, N. Bergman, U. Forssell, J. Jansson, R. Karlsson, P.-J. Nordlund, "Particle filters for positioning, navigation, and tracking", IEEE Transactions on Signal Processing, Feb 2002,Volume: 50,Issue: 2pp 425- 437
[22]C. Yang ,R. Duraiswami ,L. Davis ,"Fast Multiple Object Tracking via a Hierarchical Particle Filter", Tenth IEEE International Conference on Computer Vision.,2005,pp212-219 Vol. 1
[23]G. Bradski and A. Kaehler,”Learning OpenCV”,O'Reilly Media,2008.
[24]ADI ADXL345 Datasheet , http://www.analog.com/en/index.html
[25]STM L3G4200 Datasheet, http://www.st.com/internet/com/home/home.jsp
[26]劉瑞禎、于仕琪(2009)，OpenCV 教程，北京航空航天大學出版社
[27]劉瑞禎、于仕琪(2009)，學習OpenCV，清華大學出版社
[28]吳上立、林明德(2010)，C 語言數位影像處理-Windows/X-Window，全華圖書公司
[29]鄭明、鄭世偉(2002)，C++ Builder & Windows API，文魁資訊公司
[30]范逸之、江文賢、陳立元 (2002)C++ Builder 與 RS-232 串列通訊控制，文魁資訊公司
[31]蕭富貴(2003)，Borland C++ Builder 6實用教學寶典，台科大圖書
[32]陳錦輝(2003)，C++ Builder 6 完全攻略，金禾資訊公司