Adaptive Real-Time Decomposition of Electromyogram During Sustained Muscle Activation: A Simulation Study

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Real-time decomposition of electromyogram (EMG) signals into motor unit (MU) activity is vital for neurophysiology and human-machine interfaces.
• Existing methods struggle with signal variations like amplitude drifts and MU recruitment changes during long recordings.

Purpose of the Study:

• To develop an adaptive real-time EMG decomposition algorithm for sustained muscle activation.
• To address limitations of current methods in handling signal variability over time.

Main Methods:

• A parallel-double-thread computation algorithm was developed.
• A backend thread refines MU information using independent component analysis and convolution kernel compensation.
• A frontend thread performs real-time decomposition, evaluated on synthesized EMG data with simulated variations.

Main Results:

• The adaptive algorithm increased identifiable MU numbers by 3-4 fold over 30 minutes compared to non-adaptive methods.
• Decomposition accuracy improved by up to 10%, particularly at higher muscle contraction levels and signal-to-noise ratios.
• Performance was robust across simulated signal variations and noise levels.

Conclusions:

• The adaptive algorithm maintains decomposition performance over extended recording periods.
• It enables continuous tracking of individual MUs during sustained muscle activation.
• The approach facilitates longitudinal evaluation of MU properties and enhances neural decoding for machine interactions.