Classification of single trial EEG during imagined hand movement by rhythmic component extraction

Area of Science:

• Neuroscience
• Signal Processing
• Machine Learning

Background:

• Brain-computer interfaces (BCIs) require efficient EEG signal feature extraction.
• Rhythmic Component Extraction (RCE) is a novel method for multi-channel EEG feature extraction.
• RCE isolates specific frequency components from sensor signals.

Purpose of the Study:

• To apply RCE for extracting features related to hand movement imagery from EEG signals.
• To evaluate the classification accuracy of RCE-extracted features using machine learning.
• To compare the performance of RCE against Common Spatial Patterns (CSP).

Main Methods:

• Applied RCE to multi-channel EEG data from hand movement imagery tasks.
• Classified single-trial EEG features using machine learning algorithms.
• Compared RCE combined with Fisher Discriminant Analysis and other classifiers against CSP.

Main Results:

• RCE combined with Fisher Discriminant Analysis achieved higher classification accuracy than CSP for two subjects.
• RCE demonstrated superior performance with other major classifiers compared to CSP.
• EEG classification accuracy decreased with shorter data lengths.

Conclusions:

• RCE is an effective method for extracting discriminative features in EEG for BCI applications.
• RCE offers improved performance over CSP for classifying hand movement imagery.
• Data length is a critical factor influencing BCI classification accuracy.