Distance-Structured Lag-Sparse Graph Convolution with a Lag-Shared Adjacency Residual for Imagined Speech EEG Decoding
Under review
Research approach
Models imagined-speech EEG with distance-specific graph supports and residual adjacency shared across temporal lags.
Results
Matched subject-dependent accuracy: 84.33 ± 4.59% on BCI Competition Track 3 and 41.08 ± 2.60% on Thinking Out Loud.
All publications