An Adaptive EEG Classification Algorithm Based on CSSD and ELM_Kernel for Small Training Samples.

Area of Science:

• Neuroscience and Biomedical Engineering
• Rehabilitation Technology

Background:

• Brain-computer interfaces (BCI) offer promising rehabilitation for patients with dyskinesia, enabling movement restoration.
• BCI systems typically require a calibration phase, which can be time-consuming, especially with limited training data.
• Efficient classification of electroencephalogram (EEG) signals is crucial for reducing BCI training time.

Purpose of the Study:

• To develop a novel method for classifying EEG signals with a small amount of training data.
• To reduce the initial training time required for BCI applications in rehabilitation.
• To improve the accuracy and efficiency of motor imagery EEG signal classification.

Main Methods:

• Proposed a novel combination of feature extraction and classification algorithms for EEG signal classification.
• Employed a relative distance criterion for optimal channel selection after pre-processing motor imagery EEG signals.
• Utilized Common Spatial Subspace Decomposition (CSSD) and Extreme Learning Machine with Kernel (ELM_Kernel) for classification.

Main Results:

• Achieved a high average classification accuracy of 99.1% on the BCI Competition III dataset IVa.
• Obtained a classification accuracy of 76.92% on the BCI Competition IV dataset IIa.
• The proposed method outperformed existing state-of-the-art algorithms in classifying motor imagery EEG signals.

Conclusions:

• The novel feature extraction and classification method effectively classifies EEG signals with limited training data.
• This approach significantly reduces the calibration time needed for BCI systems.
• The findings support the potential of this method for enhancing BCI-based rehabilitation technologies.