Combining Improved Gray-Level Co-Occurrence Matrix With High Density Grid for Myoelectric Control Robustness to...

Area of Science:

• Biomedical Engineering
• Rehabilitation Engineering
• Signal Processing

Background:

• Myoelectric control systems are susceptible to electrode shift, impacting prosthesis performance.
• Robust pattern recognition is crucial for reliable myoelectric control in daily use.

Purpose of the Study:

• To develop a novel feature, iGLCM, to enhance the robustness of myoelectric control against electrode shift.
• To evaluate the effectiveness of iGLCM using high-density electrodes and improved discrete Fourier transform (iDFT).

Main Methods:

• Utilized Gray-Level Co-occurrence Matrix (GLCM) to represent spatial distribution among high-density (HD) electrodes.
• Proposed and optimized an improved GLCM (iGLCM) feature by evaluating quantization levels and input data (iDFT).
• Compared iGLCM performance against existing robust methods under simulated electrode shift conditions.

Main Results:

• iGLCM with iDFT input and quantization level 8 demonstrated comparable accuracy without shift and significantly reduced sensitivity to 1 cm electrode shift (p < 0.05).
• The method effectively halved the impact of perpendicular shifts, a primary cause of performance degradation.
• Feature space analysis indicated iGLCM discards shift-sensitive information while retaining useful data for robust control.

Conclusions:

• iGLCM offers a robust solution for myoelectric control, mitigating performance degradation caused by electrode shift.
• High-density electrodes are vital for robust pattern recognition-based myoelectric control.
• This approach facilitates the development of more reliable and practical prosthetic control systems.