A comparative study of surface EMG classification by fuzzy relevance vector machine and fuzzy support vector machine

Area of Science:

• Biomedical Engineering
• Machine Learning
• Signal Processing

Background:

• Surface electromyographic (sEMG) signals are crucial for understanding hand motions.
• Traditional relevance vector machines (RVM) face challenges in multiclass classification.
• Support vector machines (SVM) have limitations that RVM aims to address.

Purpose of the Study:

• To introduce and evaluate a multiclass fuzzy relevance vector machine (FRVM) for sEMG-based hand motion classification.
• To address the unclassifiable regions issue in multiclass RVM problems.
• To compare FRVM performance against fuzzy support vector machine (FSVM).

Main Methods:

• Developed two fuzzy membership function-based FRVM algorithms.
• Extracted AR model coefficients (AR-RMS) and wavelet transform (WT) features from sEMG signals.
• Conducted experiments on healthy subjects and amputees performing six hand motions.

Main Results:

• FRVM achieved comparable classification accuracy to FSVM with significantly fewer support vectors.
• FRVM demonstrated substantially lower processing delay compared to FSVM.
• FSVM exhibited faster training times than FRVM.

Conclusions:

• FRVM classifiers trained with sufficient data show comparable generalization to FSVM.
• FRVM offers significant sparsity, making it suitable for real-time sEMG control.
• The proposed FRVM approach enhances multiclass classification of hand motions using sEMG signals.