[Electroencephalogram Feature Selection Based on Correlation Coefficient Analysis].

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain-computer interface (BCI) systems require accurate classification of motor imagery signals.
• Limited training data poses a significant challenge for developing effective BCI systems.
• Existing feature selection methods may not be optimal for small datasets.

Purpose of the Study:

• To propose an automated method for selecting characteristic parameters to improve BCI classification accuracy.
• To enhance the performance of BCI systems using limited motor imagery training data.
• To evaluate the efficacy of correlation coefficient analysis for feature selection in BCI.

Main Methods:

• Utilized short-time Fourier transform (STFT) for initial signal processing.
• Applied correlation coefficient analysis for automatic characteristic parameter selection.
• Employed common spatial pattern (CSP) for feature extraction and linear discriminant analysis (LDA) for classification.

Main Results:

• The proposed correlation coefficient feature selection method significantly improved classification accuracy compared to methods without it.
• The method demonstrated superior performance in selecting parameters for improved classification accuracy.
• Results indicated that correlation coefficient analysis outperformed support vector machine (SVM) optimization for feature selection in this context.

Conclusions:

• Correlation coefficient analysis is an effective method for selecting characteristic parameters in BCI systems with limited data.
• The proposed approach enhances classification accuracy, contributing to more robust BCI development.
• This method offers a valuable alternative for optimizing BCI performance, especially when training data is scarce.