Enhancing robustness of spatial filters in motor imagery based brain-computer interface via temporal learning

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Spatial filtering is crucial for feature extraction in motor imagery-based brain-computer interface (MI-BCI) EEG decoding.
• While temporal filtering is recognized for discriminative feature extraction, stabilizing these features from spatial filtering remains an underexplored area.
• Existing efforts primarily focus on external feature extraction optimization, neglecting the internal stability of spatial filtering outputs.

Purpose of the Study:

• To develop a novel approach for enhancing the robustness of temporal features in MI-BCI.
• To minimize temporal instability within extracted features to improve EEG decoding performance.
• To provide a stable foundation for future advancements in brain-computer interface technology.

Main Methods:

• Proposed a method to improve temporal feature robustness by minimizing temporal domain instability.
• Utilized Jensen-Shannon divergence to quantify temporal instability.
• Integrated decision variables into an objective function to minimize instability, enhancing the stability of feature variance and mean values.

Main Results:

• The proposed method was applied to spatial filtering models and validated on public and self-collected datasets.
• Demonstrated significant improvements in classification accuracy by enhancing temporal feature stability.
• Achieved state-of-the-art accuracy: 92.43% on BCI competition III IVa, 84.45% on BCI competition IV 2a, and 73.18% on a self-collected dataset.

Conclusions:

• Enhancing temporal feature stability significantly improves MI-BCI performance.
• The developed method offers a stable foundation for future EEG decoding advancements.
• The proposed approach shows considerable potential for practical applications in EEG decoding.