Epoch length and autoregressive-order selection for electromyography signals.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Neuroscience

Background:

• Electromyography (EMG) signal analysis is vital for understanding neuromuscular function.
• Autoregressive (AR) models are commonly used for EMG signal processing.
• The selection of appropriate model orders is critical for accurate AR model-based EMG analysis.

Purpose of the Study:

• To investigate the impact of varying electromyography (EMG) epoch lengths on the selection of autoregressive (AR) model orders.
• To determine optimal AR model orders for different EMG epoch durations.

Main Methods:

• EMG signals were segmented into epochs of 25ms, 50ms, 100ms, 250ms, and 500ms.
• Least-square errors of AR models were analyzed.
• Order-selection criteria, including Bayesian Information Criterion (BIC) and Minimum Description Length (MDL), were applied.

Main Results:

• The Bayesian Information Criterion and Minimum Description Length criteria suggest that AR models with orders below 25 are sufficient for representing needle-EMG signals from normal subjects sampled at 25kHz using 500ms epochs.
• Shorter EMG epochs necessitate the use of even smaller AR model orders for accurate signal representation.

Conclusions:

• EMG epoch length significantly influences the optimal order selection for AR models.
• Shorter EMG epochs allow for the use of simpler AR models (lower orders), potentially improving computational efficiency without sacrificing accuracy.
• These findings provide guidance for optimizing AR model-based EMG analysis based on epoch length selection.