Analysis of electrode shift effects on wavelet features embedded in a myoelectric pattern recognition system.

Area of Science:

• Biomedical Engineering
• Rehabilitation Engineering
• Signal Processing

Background:

• Myoelectric pattern recognition systems translate muscle signals into prosthesis commands but suffer from low long-term robustness.
• Socket misalignment causes electrode shifts, degrading myoelectric signal (MES) quality and classification performance.

Purpose of the Study:

• To evaluate the impact of electrode shift disturbances on wavelet features extracted from MES.
• To investigate principal component analysis (PCA) frameworks for reducing wavelet feature sets.
• To assess the adaptability of wavelet features to variability in MES due to electrode shifts.

Main Methods:

• MES were recorded from seven able-bodied subjects and one subject with transradial limb loss.
• Electrode shifts were artificially simulated across six recording sessions per subject.
• Classification accuracy was measured with and without simulated electrode shifts.
• Two PCA frameworks were applied to reduce the wavelet feature set.

Main Results:

• Classification accuracy showed a minor drop from 93.8% (no disturbances) to 88.3% (with disturbances), indicating wavelet features adapt well.
• Reduced feature sets using PCA performed significantly lower than the full wavelet feature set.
• Wavelet features demonstrated resilience to MES variability caused by electrode shifts.

Conclusions:

• Wavelet features exhibit robustness against electrode shift disturbances in myoelectric control.
• The use of wavelet features can potentially mitigate performance degradation in pattern recognition systems.
• Further research into feature set reduction methods is needed to maintain high classification performance.