Scholarlink

<p>Conventional prosthetic hands are chiefly decorative and are plagued by issues such as incomplete functionality and unwieldy mechanisms. This renders them incapable of meeting the intricate control requirements of their wearers. In this paper, a prosthetic hand has been designed to satisfy the fundamental grasping requirements of the users. A long short-term memory (LSTM) model is implemented to facilitate the prediction and identification of forearm surface electromyography (sEMG) signals. In order to enhance the accuracy of prosthetic hand control, the sEMG signals are processed. Wavelet threshold denoising is applied to reduce noise in the EMG signals. A sliding window approach is employed to optimize the time-domain (TD) and frequency-domain (FD) features, ensuring temporal consistency in characterizing hand movement intentions. A LSTM model for predicting finger bending angles is established, and this model is trained using sEMG signals. Prosthetic control experiments and prosthetic grasping experiments are conducted on the basis of forearm sEMG signals. Experimental results demonstrate the implementation of finger flexion control and object grasping functionality. This serves to verify the feasibility of the LSTM model in sEMG signal pattern recognition algorithms for prosthetic hand.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

LSTM-based regression for continuous control of an EMG prosthetic hand: performance evaluation and experimental studies

  • 期刊论文

  • 发表日期：

    2026年5月30日

查看全文
  • Songqing Zhu,
  • Xuanwei Huang,
  • SongLi Sun,
  • Yali Han,
  • Tian Yao,
  • Lugang Cao

摘要

Conventional prosthetic hands are chiefly decorative and are plagued by issues such as incomplete functionality and unwieldy mechanisms. This renders them incapable of meeting the intricate control requirements of their wearers. In this paper, a prosthetic hand has been designed to satisfy the fundamental grasping requirements of the users. A long short-term memory (LSTM) model is implemented to facilitate the prediction and identification of forearm surface electromyography (sEMG) signals. In order to enhance the accuracy of prosthetic hand control, the sEMG signals are processed. Wavelet threshold denoising is applied to reduce noise in the EMG signals. A sliding window approach is employed to optimize the time-domain (TD) and frequency-domain (FD) features, ensuring temporal consistency in characterizing hand movement intentions. A LSTM model for predicting finger bending angles is established, and this model is trained using sEMG signals. Prosthetic control experiments and prosthetic grasping experiments are conducted on the basis of forearm sEMG signals. Experimental results demonstrate the implementation of finger flexion control and object grasping functionality. This serves to verify the feasibility of the LSTM model in sEMG signal pattern recognition algorithms for prosthetic hand.