Joint MVMD-based optimal feature selection and FW-LS-TWSVM for motor imagery recognition

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Motor Imagery-Brain Computer Interface (MI-BCI) systems are crucial for neurorehabilitation and human-machine control.
• Current MI-BCI systems face limitations in decoding accuracy and real-time performance.

Purpose of the Study:

• To enhance MI-BCI decoding accuracy and real-time performance.
• To introduce a novel model combining MVMD-based feature selection and FW-LS-TWSVM for MI-BCI applications.

Main Methods:

• Multivariate Variational Mode Decomposition (MVMD) for data decomposition into Intrinsic Mode Functions (IMFs).
• Common Spatial Pattern (CSP) and F-statistic-based feature selection for optimal IMF and feature extraction.
• Application of Fuzzy Weighted Least Squares Twin Support Vector Machine (FW-LS-TWSVM) for improved outlier identification in EEG decoding.

Main Results:

• The proposed method achieved high decoding accuracies of 87.40% and 88.48% on two public datasets.
• Both MVMD-based decomposition and FW-LS-TWSVM significantly contributed to improved decoding accuracy.
• The novel approach demonstrated higher accuracy and reduced training time compared to traditional methods.

Conclusions:

• The proposed model offers a significant advancement for MI-BCI systems.
• This approach can enhance motor neurorehabilitation and human-machine interaction.
• Potential to improve the quality of life for individuals with neurological disabilities.