Related Experiment Video
Updated: Jun 21, 2025
07:35
Selecting Multiple Biomarker Subsets with Similarly Effective Binary Classification Performances
Published on: October 11, 2018
7.5K
A wrapper framework for feature selection and ELM weights optimization for FMG-based sign recognition
S Al-Hammouri1, R Barioul2, K Lweesy3
1Biomedical Engineering Department, College of Engineering, Jordan University of Science and Technology, P.O. Box 3030, Irbid 22110, Jordan.
Computers in Biology and Medicine
|July 14, 2024
Summary
This study demonstrates that optimizing Extreme Learning Machine (ELM) with a hybrid binary grey wolf particle swarm optimizer (BGWOPSO) significantly improves hand gesture recognition accuracy using force myography (FMG) sensors.
Area of Science:
- Biomedical Engineering
- Machine Learning
- Human-Computer Interaction
Background:
- Force myography (FMG) offers non-contact, high-accuracy gesture recognition.
- Classifying American Sign Language gestures is crucial for assistive technologies.
Purpose of the Study:
- To evaluate a six-sensor FMG bracelet for classifying numerous hand gestures.
- To optimize the Extreme Learning Machine (ELM) classifier using swarm intelligence algorithms for feature selection.
Main Methods:
- Investigated binary grey wolf optimizer (BGWO), binary grasshopper optimizer (BGOA), and binary hybrid grey wolf particle swarm optimizer (BGWOPSO).
- Employed BGWOPSO for feature selection and ELM optimization, a novel application.
- Collected data from multiple volunteers performing 37 distinct gestures.
Main Results:
- The BGWOPSO algorithm demonstrated superior performance in optimizing ELM.
- Feature selection and ELM optimization enhanced classification accuracy from 32% to 69.84%.
- The six-sensor FMG bracelet effectively classified a wide range of gestures.
Conclusions:
- Hybrid swarm intelligence optimization (BGWOPSO) is effective for FMG-based gesture recognition.
- Optimized ELM with BGWOPSO significantly boosts classification accuracy for American Sign Language gestures.
- A simple six-sensor FMG system can achieve robust gesture classification.
