Upper Limb Movement Classification Via Electromyographic Signals and an Enhanced Probabilistic Network

Area of Science:

• Biomedical Engineering
• Neurorehabilitation
• Signal Processing

Background:

• Surface electromyography (sEMG) signals are complex and challenging for post-stroke motor assessment.
• Advances in signal processing and machine learning offer new possibilities for analyzing non-linear sEMG data.
• Current methods for motor assessment post-stroke may not fully capture subtle movement changes.

Purpose of the Study:

• To develop and evaluate a method for identifying upper limb movements from sEMG signals in stroke patients.
• To explore the utility of sEMG for monitoring motor rehabilitation progress.
• To compare the performance of an enhanced probabilistic neural network (EPNN) against other machine learning algorithms for sEMG-based movement classification.

Main Methods:

• Utilized digital signal processing, specifically discrete wavelet transform, to analyze sEMG signals.
• Employed an enhanced probabilistic neural network (EPNN) for movement classification.
• Input sEMG data from specific movements of the Arm Motor Ability Test (AMAT) into the classification algorithm.
• Identified a key frequency domain feature: the ratio of mean absolute values between sub-bands.

Main Results:

• Achieved an average classification accuracy of 75.5%, with a maximum accuracy of 100% for upper limb movement identification.
• The EPNN method demonstrated superior performance compared to support vector machine (SVM), k-Nearest Neighbors (k-NN), and standard probabilistic neural network (PNN).
• Highlighted the significance of a specific frequency domain feature for accurate classification.

Conclusions:

• The proposed method effectively identifies and distinguishes functional upper limb movements using sEMG signals in stroke patients.
• sEMG, analyzed with advanced signal processing and machine learning (EPNN), shows significant potential for monitoring stroke motor rehabilitation.
• EPNN offers a robust and accurate approach for sEMG-based motor assessment in neurorehabilitation.