Related Experiment Video
Updated: Jan 12, 2026
Capturing Dynamic Finger Gesturing with High-resolution Surface Electromyography and Computer Vision
Published on: March 28, 2025
Enhancing gesture recognition for assisting visually impaired persons using deep learning in an IoT environment-based
Hanan Abdullah Mengash1, Basma S Alqadi2, Radwa Marzouk3,4
1Department of Information Systems, College of Computer and Information Sciences, Princess Nourah Bint Abdulrahman University, P.O. Box 84428, 11671, Riyadh, Saudi Arabia. hamengash@pnu.edu.sa.
This study introduces an advanced gesture recognition system using deep learning and an improved snake optimization algorithm to aid visually impaired individuals. The novel approach achieves 98.62% accuracy, significantly improving real-time gesture interpretation.
Area of Science:
- Computer Science
- Artificial Intelligence
- Human-Computer Interaction
Background:
- Gesture recognition (GR) is crucial for interfaces but faces challenges for visually impaired users.
- Conventional machine learning (ML) struggles with real-time performance, necessitating advanced solutions.
- Deep learning (DL) offers superior capabilities for complex pattern recognition in GR.
Purpose of the Study:
- To develop an enhanced gesture recognition system (EGRVI-DLISOA) for the visually impaired in an IoT environment.
- To leverage deep learning and an improved snake optimization algorithm for accurate real-time gesture interpretation.
- To address the challenges faced by visually impaired individuals in daily tasks and technology interaction.
Main Methods:
- Utilized the Sobel filter (SF) for noise elimination in gesture data.
- Employed the SqueezeNet model for efficient feature extraction from visual data.
- Implemented Long Short-Term Memory (LSTM) for gesture classification, optimized by an improved snake optimization algorithm (ISOA).
Main Results:
- The EGRVI-DLISOA technique demonstrated superior performance in gesture recognition.
- Achieved a high accuracy rate of 98.62% on the hand gestures dataset.
- Outperformed existing gesture recognition models in experimental evaluations.
Conclusions:
- The EGRVI-DLISOA approach significantly enhances gesture recognition for visually impaired users.
- Deep learning combined with optimization algorithms offers a powerful solution for real-time assistive technologies.
- This system provides a promising advancement in computer interfaces for individuals with visual impairments.
