Related Experiment Video
Updated: Jun 5, 2025
08:45
Mapping Cortical Dynamics Using Simultaneous MEG/EEG and Anatomically-constrained Minimum-norm Estimates: an Auditory Attention Example
Published on: October 24, 2012
14.6K
Delayed knowledge transfer: Cross-modal knowledge transfer from delayed stimulus to EEG for continuous attention
Pengfei Sun1, Jorg De Winne1, Malu Zhang2
1WAVES Research Group, Department of Information Technology, Ghent University, Gent, Belgium.
Summary
This study introduces a novel framework for decoding brain signals, improving brain-computer interface accuracy by 3% using audiovisual stimuli and accounting for neural response delays.
Area of Science:
- Neuroscience
- Machine Learning
- Signal Processing
Background:
- Decoding electroencephalography (EEG) signals for machine-to-human interaction faces challenges in signal representation.
- Existing methods struggle to effectively model the inherent delays in neural responses to stimuli.
Purpose of the Study:
- To introduce a novel Delayed Knowledge Transfer (DKT) framework for attention detection using spiking neurons.
- To improve the performance of brain-computer interface (BCI) systems by effectively representing EEG signals.
- To present the WithMeAttention multimodal dataset for research on distinguishing target and distractor responses.
Main Methods:
- Developed a Delayed Knowledge Transfer (DKT) framework utilizing spiking neurons for attention detection.
- Extracted patterns from audiovisual stimuli to model brain responses in EEG signals, incorporating response delays.
- Aligned audiovisual features with EEG signals in a shared embedding space.
- Utilized the WithMeAttention dataset for evaluation.
Main Results:
- Achieved a 3% improvement in accuracy on the WithMeAttention dataset compared to a baseline model.
- Demonstrated that rhythmic enhancement of visual information optimizes multi-sensory processing.
- Identified superior performance in conditions with rhythmic target presentation, with or without auditory cues.
- Confirmed that the delay layer effectively emulates neural processing delays.
Conclusions:
- The DKT framework significantly enhances BCI performance by effectively modeling EEG signals and neural delays.
- The WithMeAttention dataset provides a valuable resource for multimodal BCI research.
- Optimizing sensory information processing through rhythmic presentation and accounting for neural delays are key for advanced human-machine interaction.
