Multi-branch deep learning neural network prediction model for the development of angular biosensors based on sEMG DOI Creative Commons
Liman Yang, Zhijun Shi,

Ruming Jia

et al.

Frontiers in Bioengineering and Biotechnology, Journal Year: 2024, Volume and Issue: 12

Published: Oct. 11, 2024

Introduction Human gait motion intention recognition is very important for the lower extremity exoskeleton robot to accurately synchronize and respond user’s natural motion. And generally performed through sEMG. Deep learning neural networks perform well in dealing with high-dimensional data nonlinear relationships such as sEMG, but different deep have their own advantages types of data. Therefore, a multi-branch network, which enables process feature items, could achieve more accurate efficient recognition. The purpose this study 1) Establish network model effective estimation joint angles. 2) Quantify performance angle prediction using Methodology This involved collection sEMG plantar pressure during walking human subjects. Firstly, collected signals are filtered denoised ensure quality reliability Calculate time domain features frequency capture key information gait. Then, sensitivity difference structural data, developed, extracted used input model. output includes cycle angle, so realize angle. Results results show that proposed method has high accuracy identifying estimating successfully integrates time-domain frequency-domain provides reliable highest 95.42%, lowest 90.11%, average 92.16%. error 3.19. Discussion designed limb recognition.The can be integrated into sensor design angular biosensors, predict real time.

Language: Английский

Motoneuron-driven computational muscle modelling with motor unit resolution and subject-specific musculoskeletal anatomy DOI Creative Commons
Arnault H. Caillet, Andrew Phillips, Dario Farina

et al.

PLoS Computational Biology, Journal Year: 2023, Volume and Issue: 19(12), P. e1011606 - e1011606

Published: Dec. 7, 2023

The computational simulation of human voluntary muscle contraction is possible with EMG-driven Hill-type models whole muscles. Despite impactful applications in numerous fields, the neuromechanical information and physiological accuracy such provide remain limited because multiscale simplifications that limit comprehensive description internal dynamics during contraction. We addressed this limitation by developing a novel motoneuron-driven neuromuscular model, describes force-generating population individual motor units, each which was described actuator controlled dedicated experimentally derived motoneuronal control. In forward contraction, model transforms vector motoneuron spike trains decoded from high-density EMG signals into unit forces sum predicted force. control provides separate descriptions recruitment discharge decodes subject's intention. subject-specific, muscle-specific, includes an advanced activation dynamics, validated against experimental Accurate force predictions were obtained when neural controls representative activity complete pool. This achieved large dense grids electrodes medium-force contractions or methods physiologically estimate units not identified experimentally. advances state-of-the-art modelling, bringing together fields musculoskeletal finding human-machine interfacing research.

Language: Английский

Citations

14
Efficient Deep Learning Model for Analyzing Muscle Activity Patterns in Biomechanical Simulations DOI
Dharmendra Dangi, Dheeraj Kumar Dixit, Amit Bhagat

et al.

SN Computer Science, Journal Year: 2025, Volume and Issue: 6(2)

Published: Feb. 5, 2025

Language: Английский

Citations

0
Unlocking the full potential of high‐density surface EMG: novel non‐invasive high‐yield motor unit decomposition DOI Creative Commons
Agnese Grison, Irene Méndez Guerra, Alexander Kenneth Clarke

et al.

The Journal of Physiology, Journal Year: 2025, Volume and Issue: unknown

Published: March 17, 2025

Abstract The decomposition of high‐density surface electromyography (HD‐sEMG) signals into motor unit discharge patterns has become a powerful tool for investigating the neural control movement, providing insights neuron recruitment and behaviour. However, current algorithms, while effective under certain conditions, face significant challenges in complex scenarios, as their accuracy yield are highly dependent on anatomical differences among individuals. To address this issue, we recently introduced Swarm‐Contrastive Decomposition (SCD), which dynamically adjusts contrast function based distribution data. Here, demonstrate ability SCD identifying low‐amplitude action potentials effectively handling scenarios. We validated using simulated experimental HD‐sEMG recordings compared it with state‐of‐the‐art methods varying including different excitation levels, noise intensities, force profiles, sexes muscle groups. proposed method consistently outperformed existing techniques both quantity decoded units precision firing time identification. Across detected, average, 25.9 ±5.8 vs . 13.9 ± 2.7 found by baseline approach. detected 19.8 13.5 units, to 11.9 6.9 method. In conditions high synchronisation approximately three times many previous (31.2 4.3 SCD, 10.5 1.7 baseline), also significantly improving accuracy. These advancements represent step forward non‐invasive EMG technology studying activity image Key points High‐density provides information how nervous system controls muscles, but struggle conditions. (SCD) is new approach that separated, increasing sample units. successfully identifies more those signals, performs well even challenging such high‐interference signals. ballistic contractions, than could improve studies fatigue neurological disorders.

Language: Английский

Citations

0
MS-MANO: Enabling Hand Pose Tracking with Biomechanical Constraints DOI
Pengfei Xie, Wenqiang Xu, Tutian Tang

et al.

2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Journal Year: 2024, Volume and Issue: 22, P. 2382 - 2392

Published: June 16, 2024

Language: Английский

Citations

3
Adaptive EMG decomposition in dynamic conditions based on online learning metrics with tunable hyperparameters DOI Creative Commons
Irene Méndez Guerra, Deren Y. Barsakcioglu, Dario Farina

et al.

Journal of Neural Engineering, Journal Year: 2024, Volume and Issue: 21(4), P. 046023 - 046023

Published: July 3, 2024

. Developing neural decoders robust to non-stationary conditions is essential ensure their long-term accuracy and stability. This particularly important when decoding the drive muscles during dynamic contractions, which pose significant challenges for stationary decoders.

Language: Английский

Citations

0
Multi-branch deep learning neural network prediction model for the development of angular biosensors based on sEMG DOI Creative Commons
Liman Yang, Zhijun Shi,

Ruming Jia

et al.

Frontiers in Bioengineering and Biotechnology, Journal Year: 2024, Volume and Issue: 12

Published: Oct. 11, 2024

Introduction Human gait motion intention recognition is very important for the lower extremity exoskeleton robot to accurately synchronize and respond user’s natural motion. And generally performed through sEMG. Deep learning neural networks perform well in dealing with high-dimensional data nonlinear relationships such as sEMG, but different deep have their own advantages types of data. Therefore, a multi-branch network, which enables process feature items, could achieve more accurate efficient recognition. The purpose this study 1) Establish network model effective estimation joint angles. 2) Quantify performance angle prediction using Methodology This involved collection sEMG plantar pressure during walking human subjects. Firstly, collected signals are filtered denoised ensure quality reliability Calculate time domain features frequency capture key information gait. Then, sensitivity difference structural data, developed, extracted used input model. output includes cycle angle, so realize angle. Results results show that proposed method has high accuracy identifying estimating successfully integrates time-domain frequency-domain provides reliable highest 95.42%, lowest 90.11%, average 92.16%. error 3.19. Discussion designed limb recognition.The can be integrated into sensor design angular biosensors, predict real time.

Language: Английский

Citations

0