An adaptation strategy of using LDA classifier for EMG pattern recognition.

Area of Science:

• Biomedical Engineering
• Rehabilitation Engineering
• Signal Processing

Background:

• Surface electromyography (sEMG) signals exhibit time-varying characteristics, often leading to reduced accuracy in traditional supervised pattern recognition for myoelectric control.
• Controlling multifunctional myoelectric prostheses in dynamic environments presents challenges due to signal variability.

Purpose of the Study:

• To propose and evaluate an unsupervised adaptation strategy of linear discriminant analysis (ALDA) for improved electromyography (EMG)-based motion classification.
• To enhance the robustness and accuracy of myoelectric prostheses control in the presence of environmental changes and signal noise.

Main Methods:

• An unsupervised adaptation strategy, termed ALDA, was developed using probability weighting and cycle substitution.
• The ALDA classifier was trained and tested using surface EMG recordings from multiple motion patterns.
• Performance comparison between ALDA and traditional linear discriminant analysis (LDA) was conducted with varying levels of added noise.

Main Results:

• The proposed ALDA method demonstrated superior performance in improving the classification accuracy of sEMG pattern recognition compared to the traditional LDA method.
• ALDA achieved better classification accuracy in both stable conditions and when surface EMG recordings were subjected to added noise.
• The adaptation strategy effectively addressed the time-varying nature of myoelectric signals.

Conclusions:

• The unsupervised ALDA strategy offers a significant improvement for EMG-based motion classification, particularly in dynamic and noisy environments.
• This approach enhances the reliability and performance of myoelectric prostheses control systems.
• ALDA provides a robust solution for real-world applications requiring accurate and adaptive myoelectric signal interpretation.