SilentWhisper
Speech recognition for whispers too quiet for people nearby to hear
SilentWhisper explores voice interaction through extremely quiet whispers that are difficult for people nearby to hear. A headset microphone captures the faint speech, and a deep-learning model recognizes it without requiring training data from each new user.
The system aims to make voice interfaces more practical in shared spaces and when handling private information, helping people interact with computers without speaking at a normal conversational volume.