Spatiotemporal-Filtering-Based Channel Selection for Single-Trial EEG Classification.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain-computer interfaces (BCIs) using electroencephalogram (EEG) require numerous channels for high classification performance, limiting practical applications.
• Identifying subject-specific optimal EEG channel subsets without performance compromise remains a significant challenge.

Purpose of the Study:

• To propose a novel method, Spatiotemporal-Filtering-based Channel Selection (STECS), for automatic identification of discriminative EEG channels.
• To leverage spatiotemporal information for efficient channel selection in EEG-based BCIs.

Main Methods:

• STECS frames channel selection as a spatiotemporal filter optimization problem incorporating group sparsity constraints.
• A computationally efficient algorithm was developed to solve the STECS optimization problem.
• The method was evaluated on three motor imagery EEG datasets.

Main Results:

• STECS achieved comparable classification performance to state-of-the-art methods using only half the number of EEG channels.
• The proposed STECS method significantly outperformed existing channel selection techniques.
• The results demonstrate the effectiveness of STECS in reducing channel requirements while maintaining high performance.

Conclusions:

• STECS offers a promising approach for simplifying BCI hardware setups.
• The method facilitates the practical utility of EEG-based BCIs by reducing channel count.
• STECS effectively utilizes spatiotemporal EEG data for optimized channel selection.