查詢結果分析
相關文獻
- A New Electrode System for Hand Action Discrimination
- Neural Network Procedures for Taguchi's Dynamic Problems
- A Fast and Efficient Competitive Learning Design Algorithm Based on Weight Vector Training in Transform Domain
- 專家系統振動訊號圖型判別之研究
- 反傳遞模糊類神經網路於流量推估之應用
- 類神經網路(Neural Networks)的種類及其在影像處理上的應用
- C++Fuzzy類神經網路物件導向發展系統之建立
- 臺灣汽保費率之估計--對數線性費率模式與類神經網路之比較
- 運用類神經網路於股價指數之套利--以日經225指數為例
- 使用類神經網路預估碳化鎢材料放電加工性能
頁籤選單縮合
題 名 | A New Electrode System for Hand Action Discrimination=辨識手部動作之主動電極系統 |
---|---|
作 者 | 徐良育; 陳建宇; 田榮雯; 胡威志; | 書刊名 | Journal of Medical and Biological Engineering |
卷 期 | 22:4 2002.12[民91.12] |
頁 次 | 頁211-217 |
分類號 | 410.1644 |
關鍵詞 | 手部動作; 主動電極; 類神經網路; Hand action; Active elctrode; Artificial neural network; |
語 文 | 英文(English) |
中文摘要 | 手語主要藉由手部、手指、手腕、手上臂等部位肌肉纖維收縮、伸張所產生之不同手勢而形成,因此本研究提議建構一以肌電圖為基礎之手部動作辨識系統。此系統之目的在於辨識足夠多的基本手部動作以利後續發展為手語辨識系統。本系統利用七個主動電極以環狀排列的方式構成一腕套系統,並將電極系統置放於前臂以擷取肌電訊號。本研究採用多閥值的訊號處理方式來偵測每個動作間期以避免雜訊的干擾。每次動作的肌電訊號特徵值則輸入倒傳遞類神經網路中辨識以達到區分手部動作之目的。本研究定義了11種手部動作利用11位受測者進行測試,並設計數種不同實驗,調整特徵法的個數,以對系統辨識能力進行測試。實驗結果方面,在使用七組電極、六種特徵法的情況下,其辨識率為93.1%。若僅取其中一種特徵值時,辨識率則在73.2%及90.4%之間。又若取兩種特徵值時,辨識率可提升為86.9%及90.3%。由實驗結果可發現,本系統對手指、手腕、手掌等動作有良好的辨識效果,但因為電極皆安排置放於前臂,故本系統對手上臂的揮動不具分辨能力。 |
英文摘要 | Sign languages are composed of hand and finger actions and are combinations of the flexion and extension of fingers, wrist, forearm, and arm. Thus, an EMG-based hand action identification system is proposed in this study. The purpose of this system is to identify a sufficient amount of basic hand actions in order to use this information to recognize more complicated sign language in the future. This system uses active electrodes placed around the forearm to collect EMG signals from muscle groups of the forearm. To avoid miss-identification of the action period due to noise and artifacts in the EMG signals, in this study a multi-thresholds method is proposed. Features extracted from the new EMG electrode system are inputted to a back-propagation ANN identification system for hand action discrimination. Eleven subjects were recruited for this study. The results indicate that when six features from seven EMGs were input into the ANN, the average discrimination rate was 93.1%. When one feature from each channel was used, the discrimination rates ranged from 73.2% to 90.4%. On the other hand, when two features with the highest discrimination rate in the previous results were selected, the average discriminative rate increased to 86.9% and 90.3%. However, the current system cannot detect movements of the upper arm. Additionally, due to the large between-subject variations, the system must go through the training sequence before every use. Nevertheless, the results indicate that, with the ring electrode system and multi-thresholds method, the proposed system does provide high discriminative ability for the actions of fingers, palm, wrist, and forearm. |
本系統中英文摘要資訊取自各篇刊載內容。