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

TrustInAging: Digital Tools for Healthy Aging.

Studies in health technology and informatics·2026
Same author

Interplay between balance, gait kinematic and physical activity level in facioscapulohumeral muscular dystrophy.

Scientific reports·2025
Same author

Deep Learning Models Optimization for Gait Phase Identification from EMG Data During Exoskeleton-Assisted Walking.

Biomimetics (Basel, Switzerland)·2025
Same author

Four-Week Exoskeleton Gait Training on Balance and Mobility in Minimally Impaired Individuals with Multiple Sclerosis: A Pilot Study.

Bioengineering (Basel, Switzerland)·2025
Same author

Power Output, Lactatemia, and Maximum Oxygen Consumption During a Specific Off-Water Incremental Test in International-Level Podium-Winner Kayak and Rowing Athletes.

Journal of functional morphology and kinesiology·2025
Same author

Power-Load Relationship of Bench Press, Ballistic Bench Press, and Prone Bench Pull in International Medal-Winning Canoeists and Kayakers.

Sports (Basel, Switzerland)·2025

Related Experiment Video

Updated: Jul 27, 2025

Estimation of Contact Regions Between Hands and Objects During Human Multi-Digit Grasping
09:41

Estimation of Contact Regions Between Hands and Objects During Human Multi-Digit Grasping

Published on: April 21, 2023

1.6K

Recurrent Network Solutions for Human Posture Recognition Based on Kinect Skeletal Data.

Bruna Maria Vittoria Guerra1, Stefano Ramat1, Giorgio Beltrami1

  • 1Laboratory of Bioengineering, Department of Electrical, Computer and Biomedical Engineering, University of Pavia, 27100 Pavia, Italy.

Sensors (Basel, Switzerland)
|June 10, 2023
PubMed
Summary
This summary is machine-generated.

This study explored Recurrent Neural Networks (RNNs) for monitoring daily living postures using skeletal data from RGB-D sensors. A 3BGRU model with data augmentation achieved 88% accuracy in detecting postures and dangerous situations.

Keywords:
ambient assisted livingdeep learninghuman action recognitionrecurrent neural networkskeletal data

More Related Videos

A Human-machine-interface Integrating Low-cost Sensors with a Neuromuscular Electrical Stimulation System for Post-stroke Balance Rehabilitation
11:06

A Human-machine-interface Integrating Low-cost Sensors with a Neuromuscular Electrical Stimulation System for Post-stroke Balance Rehabilitation

Published on: April 12, 2016

10.5K
Sit-to-stand-and-walk from 120% Knee Height: A Novel Approach to Assess Dynamic Postural Control Independent of Lead-limb
08:24

Sit-to-stand-and-walk from 120% Knee Height: A Novel Approach to Assess Dynamic Postural Control Independent of Lead-limb

Published on: August 30, 2016

10.3K

Related Experiment Videos

Last Updated: Jul 27, 2025

Estimation of Contact Regions Between Hands and Objects During Human Multi-Digit Grasping
09:41

Estimation of Contact Regions Between Hands and Objects During Human Multi-Digit Grasping

Published on: April 21, 2023

1.6K
A Human-machine-interface Integrating Low-cost Sensors with a Neuromuscular Electrical Stimulation System for Post-stroke Balance Rehabilitation
11:06

A Human-machine-interface Integrating Low-cost Sensors with a Neuromuscular Electrical Stimulation System for Post-stroke Balance Rehabilitation

Published on: April 12, 2016

10.5K
Sit-to-stand-and-walk from 120% Knee Height: A Novel Approach to Assess Dynamic Postural Control Independent of Lead-limb
08:24

Sit-to-stand-and-walk from 120% Knee Height: A Novel Approach to Assess Dynamic Postural Control Independent of Lead-limb

Published on: August 30, 2016

10.3K

Area of Science:

  • Computer Science
  • Artificial Intelligence
  • Biomedical Engineering

Background:

  • Ambient Assisted Living (AAL) systems enhance daily life support for individuals, particularly the frail.
  • Privacy concerns with cameras in AAL are addressed by RGB-D devices extracting skeletal data.
  • Deep learning, specifically Recurrent Neural Networks (RNNs), shows promise for analyzing skeletal data to recognize human postures.

Purpose of the Study:

  • To evaluate the performance of two RNN models (2BLSTM and 3BGRU) for identifying daily living postures and dangerous situations.
  • To compare the effectiveness of human-crafted kinematic features versus raw skeletal joint coordinates.
  • To investigate the impact of data augmentation on model generalization for home monitoring.

Main Methods:

  • Utilized 3D skeletal data from Kinect V2 devices for posture recognition.
  • Trained and tested two RNN models: 2BLSTM and 3BGRU.
  • Compared two feature sets: 8 kinematic features and 52 joint coordinates with distance.
  • Applied data augmentation to the 3BGRU model to balance the training dataset.

Main Results:

  • The 3BGRU model achieved an accuracy of 88% when combined with data augmentation.
  • Performance varied between the two tested feature sets.
  • The study demonstrated the feasibility of using RNNs on skeletal data for AAL applications.

Conclusions:

  • The 3BGRU model with data augmentation shows significant potential for accurate posture and activity recognition in AAL.
  • Skeletal data analysis using deep learning offers a privacy-preserving approach for home monitoring.
  • Further research can optimize feature extraction and model architectures for enhanced AAL systems.