對老年人及運動傷害者而言，如果有個輔具能辨識出使用者真正的動作意圖且幫助其施力的話，將可以大大的提升他們的生活品質及獨立生活的能力。而肌電訊號是人體肌肉最簡單且最方便取得的訊號。在臨床病理研究中，肌電訊號已廣泛的應用於診斷肌肉性、神經性病變及設計控制義肢的動作等等。因此建構一可攜式的大腿肌電訊號辨識系統，作為主動式下肢輔具的控制命令輸入，為本研究的主要動機。而研究的架構上，將以數位訊號處理器實現整個濾波、特徵值的擷取以輸入類神經網路進行辨識。
本研究的目的，以倒傳遞類神經網路作為肌電訊號辨識系統的主軸，並將之實現在數位訊號處理器中。研究中，先以Matlab軟體作為軟體程式驗證的環境。首先，將由肌電訊號擷取與放大電路所取得之大腿肌電訊號經由小波轉換，取得所需的頻帶後，透過基本特徵值萃取處理，接著送進類神經網路做實際應用的模擬與測試。以軟體印證其可行性之後，將整個流程撰寫於數位控制處理器中進行辨識，以利將來實際用於下肢輔具的控制。

Supposed that there is an instrument that could identify a user's true intention to move and assist in exerting forces, it would substantially improve the quality of life and independence of the elderly and injured. The electromyogram(EMG) is that the simplest and the most convenient electrical signals that can be acquired from human muscle. On the other hand, the electromyogram(EMG) is used extensively on diagnosing muscular or nerve pathological disorder. Additionally, EMG is also used in control of prosthesis. Therefore, the objection of this study is to build up a portable EMG identification system to be the input command of an actively lower limbs prosthesis.
In this study, an artificial back-propagation neural network for EMG identification is realized on a Digital Signal Processor(DSP). First, a Matlab program is developed to test the feasibility of the software. In the beginning, the EMG would be collected by the signal-conditioning circuit. Then, a wavelet transformation was applied to preserve the desired frequency band but filtered out the others. After the basic feature attraction, the signals would be sent into an artificial neural network to proceed the simulation. Finally, the complete algorithm would be realized on a DSP chip.

[1] C. J. Luca, “The use of surface electromyography in biomechanics,” Journal of Applied Biomechanics, pp. 135-163, 1997.
[2] 陳慶昌, Recognition of EMG signal by use of Neural Network, 國立清華大學動力機械工程學系碩士論文, 1997.
[3] A. L. Hof, H. Elzinga, W. Grimmius and J. P. K. Halbertsma, “Speed dependence of averaged EMG profiles in walking,” Laboratory of Human Movement Analysis, Department of Rehabilitation, University Hospital Groningen, Groningen, The Netherlands Institute of Human Movement Science, University of Groningen, 2001.
[4] 盛啟峰, 作業場所上肢重複性傷害現場監測技術評估探討, 朝陽科技大學工業工程與管理系碩士論文, 2003.
[6] Ingrid Daubechies, ”Where do wavelets come from? A personal point of view, ” Proceeding of the IEEE, vol. 84, no. 4, pp.510-513, 1996.
[7] Ingrid Daubechies, ”Wavelet: a tool for time-frequency analysis,” Multidimensional Signal Processing Workshop, Sixth, IEEE Int. Conf , Pacific Grove, CA, pp.98, 1998.
[8] T. B. Littler, Dr. D. J. Morrow, ”Wavelets for the Analysis and Compression of Power System Disturbances,” IEEE Trans. Power Delivery, vol. 14, no 2, pp.358-364, 1999.
[9] Olivier Poisson, Pascal Rioual and Michel Meunier, ” Detection and Measurement of Power Quality Disturbances Using Wavelet Transform,” IEEE Trans. Power Delivery, vol. 15, no. 3, pp.1039-1044, 2000.
[10] Q. Zhang and A. Benveniste, ”Wavelet Networks,” IEEE Tran. Neural Networks, vol. 3, no.11, pp.889-898, 1992.
[11] Dinesh Kant Kumar, Nemuel D. Pah and Alan Bradley, ”Wavelet analysis of surface electromyography to determine muscle fatigue,” IEEE Transaction on Neural System and Rehabilitation Engineering, vol. 11, no. 4, 2003
[12] Jun-Uk Chu, Inhyuk Moon, Shin-Ki Kim and Mu-Seong Mun, “Control of multifunction myoelectric hand using a real-time EMG pattern recognition,” Laboratory Korea Orthopedics & Rehabilitation Engineering Center, 2004.
[13] 陳志宇, 光碟機無刷主軸馬達之變轉速控制器設計, 國立清華大學動力機械工程學系碩士論文, 2006.
[14] 田榮雯, 以FPGA實現道傳遞類神經網路並應用於肌電圖分類,中原大學醫學工程學系碩士論, 2001.
[15] 蔡奇男, 以小波類神經網路實現彩色影像之臉部偵測, 朝陽科技大學資訊工程學系碩士論文, 2004.
[16] Carlo J. De Luca, “Surface electromyography: Dection and recording,” DELSYS, 2002.
[17] 黃建興, DSP/CPLD 數位電路發展系統的設計與製作, 中原大學電機工程學系碩士論文, 2002.
[18] ADSP-21161 SHARC DSP Hardware Reference, 3rd edition. Norwood, MA: Analog Devices, Inc., 2002.
[19] ADSP-21161N EZ-KIT LITE Evaluation System Manual, 2nd edition. Norwood, MA: Analog Devices, Inc., 2003.