Scholarlink

<p>The continuous prediction of joint angles from surface electromyography (sEMG) signals is essential for improving human-robot interaction systems in rehabilitation and assistive robotics. However, sEMG signals are often contaminated by noise and artifacts, posing a significant challenge to accurate and real-time prediction of joint movements. This study proposes a hybrid method combining an advanced denoising algorithm, Black Widow Optimization-Flexible Analytical Wavelet Transform (BWO-FAWT), and a predictive model based on gated recurrent units (GRU). Following the application of this method to experimental data, significant improvements in signal quality and prediction accuracy were observed. The BWO-FAWT effectively optimizes signal quality with an average signal-to-noise ratio (SNR) of 15.72±1.53 dB. The GRU model achieves high prediction performance, with a root mean square error (RMSE) of <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(1.1805\pm 0.0735^{\circ }\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>1.1805</mn> <mo>±</mo> <mn>0</mn> <mo>.</mo> <msup> <mn>0735</mn> <mo>∘</mo> </msup> </mrow> </math></EquationSource> </InlineEquation>, a mean absolute error (MAE) of <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(0.943\pm 0.0646^{\circ }\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>0.943</mn> <mo>±</mo> <mn>0</mn> <mo>.</mo> <msup> <mn>0646</mn> <mo>∘</mo> </msup> </mrow> </math></EquationSource> </InlineEquation>, and a coefficient of determination (R2) of 0.9669±0.0108, with a computation time of 19 ms per prediction and a 60 ms anticipation. Consequently, the obtained results demonstrate that this approach is suitable for real-time applications. These promising results lay the groundwork for intelligent robotic rehabilitation devices and effective dynamic assistance systems.</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

Continuous ankle motion prediction from bi-channel EMG signals: a hybrid BWO-FAWT and GRU approach

  • 期刊论文

  • 发表日期：

    2025年10月25日

查看全文
  • Ali Zakaria Messaoui,
  • Mohamed Amine Alouane,
  • Mohamed Guiatni,
  • Laurent Peyrodie,
  • Enzo Fernandez,
  • Haopping Wang

摘要

The continuous prediction of joint angles from surface electromyography (sEMG) signals is essential for improving human-robot interaction systems in rehabilitation and assistive robotics. However, sEMG signals are often contaminated by noise and artifacts, posing a significant challenge to accurate and real-time prediction of joint movements. This study proposes a hybrid method combining an advanced denoising algorithm, Black Widow Optimization-Flexible Analytical Wavelet Transform (BWO-FAWT), and a predictive model based on gated recurrent units (GRU). Following the application of this method to experimental data, significant improvements in signal quality and prediction accuracy were observed. The BWO-FAWT effectively optimizes signal quality with an average signal-to-noise ratio (SNR) of 15.72±1.53 dB. The GRU model achieves high prediction performance, with a root mean square error (RMSE) of \(1.1805\pm 0.0735^{\circ }\) 1.1805 ± 0 . 0735 ∘ , a mean absolute error (MAE) of \(0.943\pm 0.0646^{\circ }\) 0.943 ± 0 . 0646 ∘ , and a coefficient of determination (R2) of 0.9669±0.0108, with a computation time of 19 ms per prediction and a 60 ms anticipation. Consequently, the obtained results demonstrate that this approach is suitable for real-time applications. These promising results lay the groundwork for intelligent robotic rehabilitation devices and effective dynamic assistance systems.