Related Experiment Video
Updated: Aug 22, 2025
08:15
Capturing Dynamic Finger Gesturing with High-resolution Surface Electromyography and Computer Vision
Published on: March 28, 2025
704
Multi-modality deep forest for hand motion recognition via fusing sEMG and acceleration signals
Yinfeng Fang1, Huiqiao Lu1, Han Liu2
1School of Communication Engineering, Hangzhou Dianzi University, Hangzhou, 310018 Zhejiang China.
International Journal of Machine Learning and Cybernetics
|November 7, 2022
Summary
This study introduces a multi-modality deep forest (MMDF) framework for hand motion recognition, fusing surface electromyographic (sEMG) and acceleration (ACC) signals. The MMDF framework achieves higher accuracy for human-machine interaction tasks.
Area of Science:
- Biomedical Engineering
- Human-Computer Interaction
- Machine Learning
Background:
- Bio-signal based hand motion recognition is crucial for human-machine interaction, particularly for controlling prosthetics.
- Current classification technologies face challenges in achieving high accuracy with multi-modality inputs.
Purpose of the Study:
- To propose a novel multi-modality deep forest (MMDF) framework for accurate hand motion recognition.
- To fuse surface electromyographic (sEMG) and acceleration (ACC) signals at the input level for improved performance.
Main Methods:
- The MMDF framework involves sEMG and ACC feature extraction, dimension reduction, and a cascade structure deep forest classifier.
- Evaluation was performed using the public "Ninapro DB7" database.
Main Results:
- The proposed MMDF framework demonstrated significantly higher accuracy compared to existing methods.
- MMDF outperformed traditional classifiers using only sEMG signals, highlighting the value of ACC data.
Conclusions:
- ACC signals serve as an effective supplement to sEMG signals for hand motion recognition.
- The MMDF framework offers a viable solution for fusing multi-modality bio-signals in human motion recognition applications.
