透過您的圖書館登入
IP:3.144.244.44
  • 學位論文

即時手部動作辨識系統之實現

Implementation of Real Time Hand Motion Identification System

指導教授 : 徐良育

摘要


本研究的目的是以先前手部動作辨識之系列研究為基礎,發展一即時手語動作辨識系統,利用辨識表面肌電圖(EMG)的方式來分類手部動作,除了能夠即時辨識手語動作,也期望能帶給聽語障人士更多的便利。本研究首先進行兩個前導研究探討不同小波拆解肌電訊號以及類神經網路簡化對於辨識率的影響。前導研究之結果顯示此兩種方法對於辨識率之影響皆不明顯,因此本研究以最簡單之架構來實現系統。 研究中所架構之即時手部動作辨識系統,利用改良後的7頻道主動電極腕帶來擷取肌電訊號,以TMS320VC5510 DSP單晶片為訊號處理核心、MSP430控制週邊裝置。在演算流程中,本研究利用integral of EMG、waveform length、Willison amplitude三種肌電訊號特徵方式來強化不同肌電訊號的特性,再使用倒傳遞類神經網路來分類肌電訊號。此系統可即時分辨十一種不同之手部動作,其辨識率可達到88.99%。另外,改良過後的電極系統,除了能更有效地擷取肌電訊號之外,在配戴上也較為舒適,適合長時間使用。由於本系統之辨識流程非常具有彈性,可依實際應用增加或減少辨識之動作,並不侷限於辨識固定種類的動作,所以不僅可運用於手語辨識亦可推廣為機械手臂控制。

並列摘要


The goal of this study is, based on the previous researches of hand motion identification, to implement a real time hand motion identification system which can classify different hand motions based on surface electromyogram (EMG) identification. We hope that this system can also bring help to the hearing impaired. In order to estimate the effect of hand motion identification using different wavelet and evolutionary programming neural network (EPNET), two pilot studies were conducted. The results of pilot studies indicate that these two methods do not significantly improve the hand motion identification. Thus, a real time system using the simplest procedure is implemented. In the system, this study use the improved 7 channels active electrodes wrist-band to obtain EMG signal and use TMS320VC5510 DSP chip as a digital signal processor, MSP430 chip as a controller. This study extract features of EMG signal such as: integral of EMG, waveform length and Willison amplitude, and use back propagation neural network to classify hand motions. The completed system can discriminate 11 different hand motions and its accuracy can reach 88.99%. Additionally, the improved electrode system performs better in EMG data acquisition and in wearing comfort. The identification procedure of this system is very flexible, so it can classify different types of hand motions according to various applications. Furthermore, it can not only identify hand motions but also can be applied in prostheses control.

參考文獻


[6] 杜翌群,”以穩態小波結合PCA及ICA辨識手部動作肌電圖之評估”,中原大學醫工研究所,中華名國九十二年七月
[31] 江智勝,“手部動作辨識語音系統之研究”,中原大學醫學醫學工程系碩士論文,民國91年7月
[30] 田榮雯,“以FPGA實現倒傳遞類神經網路並應用於肌電圖分類”,中原大學醫學醫學工程系碩士論文,民國90年6月
[12] 陳建宇,“多電極式手部動作辨識系統”,中原大學醫學醫學工程系碩士論文,民國90年6月
[22] 林豐澤,”演化式計算上篇:演化式演算法的三種理論模式”,文化大學應用數學系

被引用紀錄


蔡百原(2007)。以數位訊號處理單晶片實現可攜式即時睡意辨識系統〔碩士論文,中原大學〕。華藝線上圖書館。https://doi.org/10.6840/cycu200700854

延伸閱讀