Not a member yet? Register for full benefits!

A True Virtual Voice: First Steps




Texas Instruments unveiled one of the first steps towards a true virtual voice in early 2008. A collar forming the ability to speak without using the vocal cords.

Rather than reading brain waves directly, the system uses the naturally occurring nerve pathways to the larynx. The brain sends the same signal to the vocal cords, regardless of whether the cords actually produce sound or not - when you form words in your head, consciously thinking, the brain is continually sending the necessary signals to the voice box, even if the voice box is 'off'.

The implications are obviously staggering. If the signals are being sent regardless of activity - which they are - then it is possible to restore speech to those whose larynx is disabled, or even make telephone calls silently, on your end. Speech controlled interfaces would be controllable even if the user has no natural voice, or a slurred one.

There is a very noticeable time delay in the current system, as at the time of showing it could only recognise 150 words. Additionally, it took time - about ten seconds - for the collar to detect the nerve signals, and for the computer to match those into words, after repeated tries from the demonstration person. Waiting ten seconds every time an answer is expected in conversation would be a nightmare, but of course this is early days, and limited vocabulary. As a greater range of signals are understood, the time taken to match to those on file is lessened, and of course, the more likely a given signal is to match something, at least.

Plans are afoot to release a communication collar for ALS sufferers by the end of 2008. As the presenter points out in the above video, the technology is similar to the early days of speech recognition: Incredibly limited. Of course incredibly limited beats nothing available, hands down.

Further Reading

The audeo

Staff Comments


Untitled Document .