Masked multimodal training on sEMG and lipreading reduces word error rate by up to 14 percentage points and improves robustness to modality loss in silent speech synthesis.
Confidence- Based Self-Training for EMG-to-Speech: Leveraging Synthetic EMG for Robust Modeling,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Cross-Modal Masking for Robust Silent Speech Synthesis Using sEMG and Lipreading
Masked multimodal training on sEMG and lipreading reduces word error rate by up to 14 percentage points and improves robustness to modality loss in silent speech synthesis.