A novel approach for SEMG signal classification with adaptive local binary patterns.

Area of Science:

• Signal Processing
• Biomedical Engineering
• Pattern Recognition

Background:

• Feature extraction is crucial for pattern recognition.
• Traditional methods like Local Binary Pattern (LBP) and 1D-LBP have limitations, including sensitivity to noise and fixed neighbor positions.
• These limitations hinder their effectiveness in detecting subtle patterns in signals like SEMG.

Purpose of the Study:

• To introduce a novel adaptive local binary pattern (aLBP) for enhanced feature extraction.
• To overcome the limitations of existing LBP methods, particularly noise sensitivity and fixed neighbor configurations.
• To improve the detection of hidden patterns in time-varying signals for potential medical applications.

Main Methods:

• Developed the adaptive local binary pattern (aLBP) by incorporating adaptive neighbor selection and smoothing coefficients.
• Applied aLBP to analyze two distinct datasets, focusing on time-varying signals.
• Compared the performance of aLBP against established feature extraction techniques.

Main Results:

• The proposed aLBP method demonstrated superior accuracy compared to existing popular feature extraction approaches.
• Results showed higher accuracy than previously reported methods in the literature for the employed datasets.
• The effectiveness of aLBP in investigating SEMG signals was validated.

Conclusions:

• The adaptive local binary pattern (aLBP) is a promising feature extraction technique.
• aLBP offers improved pattern detection capabilities, especially for noisy and complex signals.
• This method has significant potential for applications in biomedical signal analysis, such as SEMG investigation.