Unsupervised Stochastic Strategies for Robust Detection of Muscle Activation Onsets in Surface Electromyogram

Area of Science:

• Biomedical Engineering
• Neuroscience
• Signal Processing

Background:

• Surface electromyographic (sEMG) signals are crucial for understanding muscle function and diagnosing neuromuscular diseases.
• Accurate detection of muscle activation onsets (MAOs) is challenging due to inherent variations in sEMG data.
• Existing methods for MAO detection often require parameter tuning and can be sensitive to data variability.

Purpose of the Study:

• To develop and evaluate unsupervised statistical approaches for precise MAO detection in sEMG.
• To compare the performance of profile likelihood maximization (PLM) and scree-plot elbow detection (SPE) against a state-of-the-art algorithm.
• To identify a robust and adaptable method for MAO estimation without parameter tuning.

Main Methods:

• Two unsupervised statistical methods, SPE and PLM, were applied to preconditioned sEMG data.
• sEMG data included both simulated signals and real-world recordings from human participants performing specific movements.
• Performance was evaluated by comparing estimated MAO times against a gold standard using mean and median errors.

Main Results:

• The PLM-Laplacian variant achieved high accuracy, with low median errors of 9 ms (simulated) and 21 ms (actual sEMG) compared to the gold standard.
• PLM-Laplacian significantly outperformed other tested algorithms, including SPE and a state-of-the-art method.
• The PLM approach demonstrated robustness by not requiring parameter tuning, enhancing its flexibility.

Conclusions:

• The PLM-Laplacian method offers a highly accurate and robust approach for detecting muscle activation onsets from sEMG data.
• This unsupervised technique provides a significant advantage over existing methods due to its adaptability and lack of parameter dependency.
• Further validation under diverse experimental conditions is ongoing to solidify its clinical and research applications.