Voiceless Speech Recognition
Giving a ‘Voice to the Voiceless’
Voice is the primary method of face-to-face communication – someone speaks, someone listens. There are situations, however, where voice-based communication is not possible – either the person cannot speak in the normal sense or the listener cannot hear what is being spoken.
Liopa is developing a communication aid, SRAVI, to assist those who have lost the ability to speak, but who can move their lips normally. The application will initially be targeted at Tracheostomy patients and will allow users to ‘speak’ a selection of pre-defined phrases.
The on-device client captures a video of person ‘speaking’ which is then uploaded to Liopa’s cloud-based Visual Speech Recognition (VSR) engine, LipRead, for processing. The correct phrase is returned and displayed/played on smartphone.
Requires only a smartphone
The mobile app is deployed on commodity smartphone/tablets, making it very cost efficient, relative to bespoke hardware-based alternatives
It learns as you go
The supported phrase list can be updated because the system can learn new words as you use it
More than 90% accurate, and self-tuning
Continued use of the service will improve phrase selection accuracy. At least 90% of the time, the correct phrase is returned when it’s spoken from the phrase list
Improving well being
The frustration of being ‘voiceless’ is alleviated, allowing the user to communicate with the people around them, improving morale and improving health outcomes
The application is currently being trialed with ICU patients at Lancashire Teaching Hospitals NHS Trust. Find out more: www.sravi.ai
Voiceless Speech Recognition also has applications in scenarios where speech is possible, but can be heard/processed. Smart factories, the key building blocks of the Industry 4.0 initiative, advocate the use of voice for the Man Machine Interface (MMI). When compared to legacy keyword/mouse interaction, the use of voice to control machinery is quicker, more reliable and allows the operator to keep their eyes and hands on the job. In these industrial settings, however, the level of noise can be such that the operator’s voice cannot be heard and audio-based voice control performs very poorly. By equipping the machine with a basic camera, control using Voiceless Speech Recognition becomes a viable option, especially as such machines are typically controlled by a predefined set of commands.
Download SRAVI brochure
Improving patient engagement and autonomy