I recall hearing about this tech a few years ago. When one “pronounces” words in their head the vocal chords are activated to a minute degree. The idea is to capture the neurological signals on their way to the vocal cords and send them to a speech processor. At the time they were only able to distinguish “yes” from “no”. A company called Ambient now has a device call Audeo. The sensor is built into a thin, easily donned neckband, making the person look like they are wearing a turtleneck collar. Audeo can currently recognize about 150 words.
Michael Callahan, co-founder of Ambient Corporation likens their progress to the early days of speech recognition. Given the acceleration of technological development we can expect a cheap commercial version of Audeo in about 5 years. Ten years ago commercial speech recognition software was expensive ($400-$1500), needed a powerful computer and required each-word-be-spoken-discreetly. Today, the average home computer can run continuous speech recognition software for $99. In fact, if you try to use discrete speech today, you’ll just confuse the program.
This new tech get surprisingly close to telepathy, although it should be noted that it cannot read thoughts (you need an fMRI for that). It takes a fair amount of concentration and can only read words that are intended to be spoken. Still, once these words are digitized they become just another packet of data, able to be transmitted or manipulated like any other. Imagine a crowded room where everyone has a cellphone and bluetooth headset. Twenty conversations could be occurring at the same time but an observer would see a room of silent, blankly staring people.
Let’s hope they can do some work on vocal nuance or in ten years we’re all going to sound like Stephen Hawking.
Posted by Mugly 