What if a computer could recognise words without hearing them? Radio-frequency sensing is being developed to turn silent speech into a hands-free input method.

The Subvocal Company is developing smart glasses with AI technology to convert what you are silently mouthing into digital text. This technology could offer an alternative way of inputting information without having to speak or type. You could write messages, take notes, and use AI tools.
This technology does not use a microphone or camera unlike a standard voice to text instead uses radio frequency (RF) sensing. This means that the user can say words without speaking out loud.
The system sends low-power radio waves to the user’s face. As they silently mouth what they want to say, their jaw and speech muscles move, causing changes to the reflected radio waves. This change is picked up by the sensor, and machine learning models can be used to understand what they are saying.
The company states that their prototype is able to understand free text silent speech at over 200 words per minute (WPM). They are also working on ways to make the hardware smaller. They are gathering thousands of hours of RF data from multiple users, which could help improve accuracy.
If this technology can become more accurate, compact, and adaptable for multiple users, it could have many applications. This could include helping people with communication, using medical equipment hands-free, and human-machine interaction. How widely this is used will depend on how accurately it can understand what people are saying without speaking.







