












摘" 要: 人體在運動過程中會產生微弱的生物電信號,其中蘊含著大量的控制信息。為了使用生物電信號中的信息控制機械臂動作,提出一種基于DTW算法的sEMG手勢識別控制系統,利用該系統對采集的原始信號進行濾波和放大。為了確定有效的sEMG,采用移動平均法對處理信號進行劃分。使用平均絕對值從數據片段中提取有效段數據,應用DTW算法將3路表面肌電信號融合,計算樣本與模型之間的相似度,實現手勢識別;再將識別后的信號通過無線模塊發送到控制指令,以控制機械臂的動作;最后,采用提出的算法并結合6種類型的手勢分類模型創建最佳特征模型。實驗測試結果表明,使用動態時間規整(DTW)算法進行手勢識別的平均準確率為93.752%,6種手勢的平均模型匹配率達到92%,實現了肌電信號對機械臂的精確控制。由此證明所提方法的手勢識別比傳統的閾值控制開關更準確。
關鍵詞: 手勢識別; DTW算法; 表面肌電圖(sEMG); 特征提取; 機械臂; 手勢檢測
中圖分類號: TN911.71?34; TP242" " " " " " " 文獻標識碼: A" " " " " " " " " " " "文章編號: 1004?373X(2025)02?0131?06
Design of sEMG gesture recognition control system based on DTW algorithm
HAN Tuanjun, LEI Dongyuan, HUANG Chaojun, LU Chao
(School of Physics amp; Telecommunications Engineering, Shanxi University of Technology, Hanzhong 723000, China)
Abstract: The human body can produce weak bioelectric signals during movement, which contains a large amount of control information. In order to use the information from bioelectric signals to control the movement of robotic arms, a sEMG (surface electromyography) gesture recognition control system based on DTW (dynamic time warping) algorithm is proposed, which can be used to filter and amplify the collected original signal. In order to determine the effective sEMG, the moving average method is used to partition the processed signal. The MAV (mean absolute value) is used to extract the valid segment data from data fragments, the DTW algorithm is used to fuse three sEMG, and the similarity between the sample and the model is calculated, so as to realize the gesture recognition. The recognized signal is used to send the control command by means of the wireless module to control the action of the robotic arm. The proposed algorithm is used to create the optimal feature model by combining with 6 gesture classification models. The experimental results show that the average accuracy of the gesture recognition using the DTW algorithm is 93.752%, and the average model matching rate of 6 gestures can reach 92%, achieving the precise control of the robotic arm by sEMG. It proves that the gesture recognition of the proposed method is more accurate than traditional threshold controlled switches.
Keywords: gesture recognition; DTW algorithm; surface electromyography; feature extraction; mechanical arm; gesture detection
0" 引" 言
肌電信號是肌肉活動狀態中產生的動作電位序列,不同序列對應不同的肌肉收縮模式,不同肌肉收縮模式對應不同的sEMG信號特征,通過對肌電信號特征差異性研究,可以將肌電信號應用于人工智能控制等領域[1?3]。本文提出了一種基于DTW算法的sEMG手勢識別控制系統,系統通過肌電傳感器采集原始肌電信號,對原始肌電信號進行濾波和放大;為了確定有效的sEMG,采用移動平均法對信號進行處理并劃分信號段。……