Real-Time Decomposition of Multi-Channel Intramuscular EMG Signals Recorded by Micro-Electrode Arrays in Humans
IEEE Transactions on Bio-Medical Engineering
|April 1, 2025
Summary
This study introduces a real-time intramuscular electromyography (iEMG) decomposition algorithm using Bayesian filtering and GPU computing. The method accurately identifies motor unit signals for advanced human-machine interfaces.
Area of Science:
- Biomedical Engineering
- Neuroscience
- Signal Processing
Background:
- Intramuscular electromyography (iEMG) decomposition is crucial for extracting motor neuron (MN) discharge timings.
- Real-time decomposition enables the development of sophisticated human-machine interfaces (HMIs).
Purpose of the Study:
- To develop and validate a multi-channel, real-time iEMG decomposition algorithm.
- To enhance computational speed for real-time processing using parallel GPU clusters.
- To assess the algorithm's accuracy in decomposing simulated and experimental iEMG signals.
Main Methods:
- A Hidden Markov Model of EMG combined with a Bayesian filter was employed for decomposition.
- A multi-channel Bayesian modeling and filtering framework was implemented using parallel GPU computation.
- A decomposed-checked channel strategy was utilized for efficient channel grouping and processing.
Main Results:
- The algorithm achieved real-time decomposition of multi-channel iEMG signals.
- Validation on simulated and experimental data (tibialis anterior, abductor digiti minimi) demonstrated high accuracy.
- An average decomposition accuracy of 90% was achieved across all tested signals.
Conclusions:
- The proposed multi-channel iEMG decomposition algorithm is effective for real-time signal processing.
- The algorithm's compatibility with implanted multi-channel electrode arrays facilitates advanced HMIs.
- High-information transfer rates are achievable for neural decoding applications.
