Classification of the mechanomyogram signal using a wavelet packet transform and singular value decomposition for...

Area of Science:

• Biomedical Engineering
• Rehabilitation Engineering
• Signal Processing

Background:

• Mechanomyogram (MMG) signals show promise for controlling powered prostheses.
• MMG signals are susceptible to noise from movement artifacts, hindering classification accuracy.
• Existing methods struggle to reliably extract robust features from noisy MMG data.

Purpose of the Study:

• To investigate a novel scheme for classifying hand motions using MMG signals.
• To address the challenge of noise interference in MMG signals for prosthetic control.
• To enhance the accuracy and reliability of multifunctional prosthetic control systems.

Main Methods:

• Proposed a feature extraction scheme integrating Wavelet Packet Transform (WPT), Singular Value Decomposition (SVD), and distance evaluation criteria.
• Utilized WPT for effective time-frequency representation of non-stationary MMG signals.
• Employed SVD and distance evaluation for optimal feature extraction and selection from MMG time-frequency matrices.

Main Results:

• The proposed method reliably differentiated four forearm and hand motions using two-channel MMG signals across 12 subjects.
• Achieved the highest average classification accuracy of 89.7%, outperforming previous time-frequency decomposition methods.
• Demonstrated the robustness of the feature extraction scheme against noise interference.

Conclusions:

• The integrated WPT-SVD feature extraction method provides robust MMG signal classification for hand motions.
• Mechanomyogram (MMG) signals can serve as a viable alternative to electromyogram (EMG) for prosthetic control.
• The proposed classification approach significantly improves the potential for advanced, multifunctional prosthetic devices.