The authors say DRDNet builds two complementary temporal views from spatial EEG features via average and max pooling, models them with a bidirectional Mamba encoder and a Transformer encoder respectively, and integrates them through time-step-level dynamic fusion followed by long short-term memory based temporal aggregation. On the Chinese character stroke handwriting imagery (CCSHI) and pinyin single-vowel handwriting imagery (SVHI) tasks, it outperformed 7 representative EEG decoding baselines, with Cohen’s kappa of 0.5968 and 0.5502.
Dual-View Network Reaches 67.74% on Handwriting-Imagery EEG
Summary
DRDNet separates spatial EEG features into two temporal views, models them with a bidirectional Mamba encoder and a Transformer, and then combines them through dynamic fusion and LSTM aggregation. On a public dataset, it reached 67.74% accuracy for imagined Chinese-character strokes and 62.51% for imagined pinyin vowels, outperforming seven EEG-decoding baselines.
Why it matters
Handwriting imagery extends non-invasive BCI research beyond simple commands, and the reported accuracy and kappa values provide a clear basis for comparing this dual-view architecture with prior methods.
Compiled by BCIwiki from public sources
Sources · 1
pubmed.ncbi.nlm.nih.gov 2026-08-19