Adaptive certainty-based classification for decomposition of EMG signals

Area of Science:

• Biomedical Engineering
• Signal Processing
• Neuroscience

Background:

• Electromyographic (EMG) signal decomposition is crucial for understanding muscle activity.
• Accurate identification of motor unit potentials (MUPs) and their firing patterns is challenging due to signal variability.
• Existing classification methods may struggle with complex EMG signal characteristics.

Purpose of the Study:

• To develop and evaluate an adaptive certainty-based supervised classification approach for EMG signal decomposition.
• To improve the accuracy and reduce variability in classifying MUPs compared to existing methods.
• To assess the classifier's performance on both synthetic and real-world EMG signals.

Main Methods:

• Developed an adaptive certainty classifier (ACC) utilizing a similarity criterion based on MUP shape and firing patterns (passive and active modes).
• Evaluated the ACC's performance using simulated EMG signals with controlled intensity and variability.
• Compared the ACC's performance against the certainty classifier (CC) using both simulated and real EMG datasets.

Main Results:

• The ACC demonstrated superior average correct classification rates (CCr) and lower mean absolute deviation (MAD) across all tested signal types.
• For simulated signals, ACC achieved CCr of 83.7% (MAD 5.8%) vs. CC's 78.3% (MAD 8.7%) for varying intensity.
• For real signals, ACC achieved CCr of 70.0% (MAD 6.3%) vs. CC's 64.9% (MAD 6.4%), effectively handling MUP shape and firing pattern variability.

Conclusions:

• The adaptive certainty classifier (ACC) effectively manages variability in MUP shape and motor unit firing patterns.
• ACC dynamically adjusts classification criteria based on EMG signal characteristics, optimizing accuracy and minimizing errors.
• The ACC's adaptability suggests broad applicability across diverse EMG signal analysis tasks.