Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Learned spatially varying microscopy model with adaptive point spread functions.

Optics express·2026
Same author

Mechanical signatures of left ventricular ejection fraction and mass index in a community cohort (ACE 1950).

The international journal of cardiovascular imaging·2026
Same author

Transformer-Based Context-Informed Incremental Learning With sDTW Alignment Unlocks Fast and Precise Regression-Based Myoelectric Control.

IEEE transactions on neural systems and rehabilitation engineering : a publication of the IEEE Engineering in Medicine and Biology Society·2026
Same author

Readout Techniques and Offset Compensation Strategies for Biomedical Resistive MEMS Sensors: A Comprehensive Review.

IEEE reviews in biomedical engineering·2025
Same author

Fully-Flexible Multifunctional Polydimethylsiloxane (PDMS) Neural Probe With a U-Turn Polyester Microchannel.

IEEE transactions on bio-medical engineering·2025
Same author

Context Informed Incremental Learning Improves Myoelectric Control Performance in Virtual Reality Object Manipulation Tasks.

Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual International Conference·2025

Related Experiment Video

Updated: Dec 25, 2025

Capturing Dynamic Finger Gesturing with High-resolution Surface Electromyography and Computer Vision
08:15

Capturing Dynamic Finger Gesturing with High-resolution Surface Electromyography and Computer Vision

Published on: March 28, 2025

1.1K

Interpreting Deep Learning Features for Myoelectric Control: A Comparison With Handcrafted Features.

Ulysse Côté-Allard1, Evan Campbell2, Angkoon Phinyomark2

  • 1Department of Computer and Electrical Engineering, Université Laval, Quebec, QC, Canada.

Frontiers in Bioengineering and Biotechnology
|March 21, 2020
PubMed
Summary

A new Adaptive Domain Adversarial Neural Network (ADANN) improves inter-subject EMG gesture recognition accuracy by 19.40%. This deep learning approach reveals complementary information between learned and handcrafted features for better myoelectric control.

Keywords:
CNNConvNetEMGGrad-CAMMAPPERdeep learningfeature extractiongesture recognition

More Related Videos

A Structured Rehabilitation Protocol for Improved Multifunctional Prosthetic Control: A Case Study
06:58

A Structured Rehabilitation Protocol for Improved Multifunctional Prosthetic Control: A Case Study

Published on: November 6, 2015

10.1K
Characterization of the Sense of Agency over the Actions of Neural-machine Interface-operated Prostheses
05:21

Characterization of the Sense of Agency over the Actions of Neural-machine Interface-operated Prostheses

Published on: January 7, 2019

8.3K

Related Experiment Videos

Last Updated: Dec 25, 2025

Capturing Dynamic Finger Gesturing with High-resolution Surface Electromyography and Computer Vision
08:15

Capturing Dynamic Finger Gesturing with High-resolution Surface Electromyography and Computer Vision

Published on: March 28, 2025

1.1K
A Structured Rehabilitation Protocol for Improved Multifunctional Prosthetic Control: A Case Study
06:58

A Structured Rehabilitation Protocol for Improved Multifunctional Prosthetic Control: A Case Study

Published on: November 6, 2015

10.1K
Characterization of the Sense of Agency over the Actions of Neural-machine Interface-operated Prostheses
05:21

Characterization of the Sense of Agency over the Actions of Neural-machine Interface-operated Prostheses

Published on: January 7, 2019

8.3K

Area of Science:

  • Biomedical Engineering
  • Machine Learning
  • Signal Processing

Background:

  • Myoelectric control systems traditionally rely on handcrafted features from electromyographic (EMG) signals for gesture recognition.
  • Deep learning offers feature learning but faces challenges with interpretability and poor inter-subject generalization due to EMG signal variability.
  • Understanding learned features and their relationship to traditional handcrafted features is crucial for advancing EMG-based gesture recognition.

Purpose of the Study:

  • To introduce a novel multi-domain learning algorithm, ADANN, to enhance inter-subject classification accuracy in EMG gesture recognition.
  • To perform the first topological data analysis of deep learned features in EMG recognition, comparing them with handcrafted features.
  • To provide insights into the information encoded by deep networks and guide the development of hybrid feature sets.

Main Methods:

  • Development and application of the Adaptive Domain Adversarial Neural Network (ADANN) for EMG signal processing.
  • Utilizing topological data analysis to characterize information encoded in ADANN-generated features.
  • Employing convolutional network visualization techniques to analyze learned feature representations.

Main Results:

  • ADANN significantly enhanced inter-subject classification accuracy by an average of 19.40% (p = 0.00004) compared to standard training.
  • Topological analysis revealed that early learned features and handcrafted features capture distinct information for gesture discrimination.
  • Later-layer learned features adopt a one-vs.-all strategy, and visualization shows a tendency to ignore high-amplitude channels, unlike handcrafted features.

Conclusions:

  • ADANN effectively improves cross-subject generalization in EMG-based gesture recognition.
  • Learned and handcrafted features encode complementary information, suggesting potential for hybrid feature sets.
  • This work provides a foundation for interpreting deep learning models in myoelectric control and developing more robust systems.