SoniSpeech, the first open-vocabulary trimodal silent speech dataset from acoustic-sensing eyewear, achieves 26.3% WER in a CTC ResNet baseline.
Main results Table 3 summarizes the performance for each configuration
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SD 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
SoniSpeech: A Large-Scale Open-Vocabulary Tri-Modal Dataset for Wearable Silent Speech Interfaces
SoniSpeech, the first open-vocabulary trimodal silent speech dataset from acoustic-sensing eyewear, achieves 26.3% WER in a CTC ResNet baseline.