Joint application of rough set-based feature reduction and Fuzzy LS-SVM classifier in motion classification

Area of Science:

• Biomedical Engineering
• Signal Processing
• Machine Learning

Background:

• Surface electromyographic (sEMG) signals are crucial for understanding human motion.
• Accurate classification of sEMG signals is essential for applications like prosthetics and human-computer interfaces.
• Existing methods often face challenges with computational complexity and classification accuracy.

Purpose of the Study:

• To develop an effective classification scheme for sEMG-based motion classification.
• To reduce computational complexity while maintaining high classification accuracy.
• To evaluate the proposed scheme against conventional methods.

Main Methods:

• Decomposition of sEMG signals using Wavelet Packet Transform (WPT) to extract energy features from sub-bands.
• Application of Rough Set Theory (RST) for feature reduction to decrease computational load.
• Implementation of a Fuzzy Least Squares Support Vector Machine (LS-SVM) for multi-class classification.

Main Results:

• The proposed RST-based feature selection effectively reduced feature set complexity without compromising accuracy.
• The Fuzzy LS-SVM classifier demonstrated high accuracy in identifying diverse motions from sEMG signals.
• Comparative experiments showed superior performance compared to Principal Component Analysis (PCA) and Neural Network (NN) based schemes.

Conclusions:

• The combination of WPT, RST, and Fuzzy LS-SVM offers a powerful approach for sEMG motion classification.
• This scheme exhibits significant potential for advanced applications requiring precise motion interpretation.
• The RST-based feature selection is a versatile component adaptable to different classifiers.