Headphones that can read your mind

Advances in speech recognition technology have made it a more viable, but not necessarily quieter, form of computer interface. To prevent the clicking of keyboards from being replaced by loud man-machine conversations, MIT researchers are developing a new system called AlterEgo.

AlterEgo allows people to talk to computers without speaking and to listen to them without using their ears. The helmet looks like the product of a poor design student. Instead of the familiar combination of headphones and microphone, the device is a bulky white plastic curve like the jaw of a strange animal that goes from the wearer's ear and bends to touch the chin.

It may sound strange, but AlterEgo is based on fairly sophisticated technology. Inside Alterego are electrodes that scan the jaw and face of the neuromuscular signals produced when the wearer thinks of verbalizing words without actually speaking them out loud; a practice called subvocalization. This allows the head to act like a microphone for a computer without making sounds. Meanwhile, a pair of bone conduction headphones transmit sounds from the computer to the user, which bypasses the usual channels and goes directly to the inner ear via the jawbones and skull bones.

According to MIT, this makes communication with a computer silent and completely private. An example of the advantages of such a system is being able to use a computer to beat an opponent by silently communicating movements to the device and surreptitiously receiving advice. More ethical use would be on the deck of an aircraft carrier, where there is normally too much noise to speak or hear. But Arnav Kapur, leader of the project, there is a more fundamental application.

"The motivation for this was to build an AI device – an intelligence enhancement device," says Kapur. “Our idea was: could we have a computer platform that merges human and machine in some way; and which resembles an internal extension of our own cognition? "

[Embed] https://www.youtube.com/watch?v=RuUSc53Xpeg [/ embed]

In other words, such an interface could allow users to do without access codes, type and, in general, hack with digital devices, which would divert them from commercial activity.

Subvocalization has been known for more than a century, but as far as computer interfaces are concerned, it is still in its infancy. To understand how to design the AlterEgo helmet, the MIT team used a set of 16 electrodes placed on different parts of a series of volunteer faces to find the best neuromuscular signals. They were then asked to subvocalize a series of words four times in succession.

The team found that only four electrodes are really needed, so a less bulky device could emerge. Currently, the algorithms used work on a vocabulary of about 20 words each for a series of computer tasks. These include multiplication issues and a chess application that allows movements to be entered using the standard alphanumeric chess nomenclature.

Once the words are sub-located, a neural network passes the data through a series of simple, layered processing nodes that look for correlations between neuromuscular signals and specific words. Usability tests on 10 subjects, who had spent 15 minutes adjusting the prototype to their neurophysiology, followed by 90 minutes used to perform computer exercises, showed 92% accuracy.

However, Kapur says this performance could be improved with more training data for the system. The hope is that by collecting more data, the system can go beyond simple commands and responses.

AB SMART HEALTH REVIEW