MIT's AlterEgo Headset Can Hear The Voices In Your Head, Literally
AlterEgo is a rather odd-looking wearable device that mounts to your head and wraps around the back of your neck. Electrodes mounted on the device are than able to identify neuromuscular signals in the jaw and face to read a human's subvocalization. Subvocalization refers to the muscular movements that are made when we almost imperceptibly say words to ourselves. Other people might not be able to pick up the subtle lip and muscular movements that we make when we mouth words to ourselves, but AlterEgo can "hear" them loud and clear.
A total of 16 electrodes are located on the device to reliably read neuromuscular signals from various facial locations (seven of those electrodes are located on either side of a person's mouth and on the jaw). The signals are then analyzed by a machine learning system that interprets them as words.
In addition, AlterEgo is able to "speak" to you using built-in bone-conducting speakers that transmit vibrations to your inner ear. So, it is in essence a two-way system. You can speak commands, which the system should [correctly] interpret, and then you receive meaningful feedback which can be used to complete tasks.
Researchers conducted usability studies, during which a subject spent nearly 2 hours training the system their neurophysiology and executing computations with the system. Once that process was completed, transcription accuracy for the test subjects dialed in at around 92 percent.
Interestingly enough, one key uses of this technology could actually be on the battlefield with special ops personnel. “There’s a lot of places where it’s not a noisy environment but a silent environment," said Thad Starner, a professor at Georgia Tech’s College of Computing. "A lot of time, special-ops folks have hand gestures, but you can’t always see those. Wouldn’t it be great to have silent-speech for communication between these folks?"
In its current iteration, AlterEgo looks rather silly attached to a subject’s face, but there's no doubt that future versions can be miniaturized, likely using fewer sensors. And then, it should be rather trivial to use such technology to carry on inconspicuous conversations with the Google Assistant, Siri and even Alexa. The future is looking bright, indeed.