SAS models device-addressed speech detection as sequential routing over interaction history and achieves F1=0.95 with audio-video fusion on proprietary multi-speaker data while running fully on-device.
Turn-taking in human commu- nication: Origins and implications for language processing
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SD 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Selective Attention System (SAS): Device-Addressed Speech Detection for Real-Time On-Device Voice AI
SAS models device-addressed speech detection as sequential routing over interaction history and achieves F1=0.95 with audio-video fusion on proprietary multi-speaker data while running fully on-device.