Researchers have created a wearable device that can read people’s minds when they use an internal voice, allowing them to control devices and ask queries without speaking.
The device, called AlterEgo, can transcribe words that wearers verbalise internally but do not say out loud, using electrodes attached to the skin.
“Our idea was: could we have a computing platform that’s more internal, that melds human and machine in some ways and that feels like an internal extension of our own cognition?” , who led the development of the system at MIT’s Media Lab.
Kapur describes the headset as an “intelligence-augmentation” or IA device, and was presented at the Association for Computing Machinery’s in Tokyo. It is worn around the jaw and chin, clipped over the top of the ear to hold it in place. Four electrodes under the white plastic device make contact with the skin and pick up the subtle neuromuscular signals that are triggered when a person verbalises internally. When someone says words inside their head, artificial intelligence within the device can match particular signals to particular words, feeding them into a computer.
The computer can then respond through the device using a bone conduction speaker that plays sound into the ear without the need for an earphone to be inserted, leaving the wearer free to hear the rest of the world at the same time. The idea is to create a outwardly silent computer interface that only the wearer of the AlterEgo device can speak to and hear.
The AlterEgo device managed an average of 92% transcription accuracy in a 10-person trial with about 15 minutes of customising to each person. That’s several percentage points below the 95%-plus accuracy rate that Google’s voice transcription service is capable of using a traditional microphone, but Kapur says the system will improve in accuracy over time. The human threshold for voice word accuracy is thought to be around 95%.
Kapur and team are currently working on collecting data to improve recognition and widen the number of words AlterEgo can detect. It can already be used to control a basic user interface such as the Roku streaming system, moving and selecting content, and can recognise numbers, play chess and perform other basic tasks.
The eventual goal is to make interfacing with AI assistants such as Google’s Assistant, Amazon’s Alexa or Apple’s Siri less embarrassing and more intimate, allowing people to communicate with them in a manner that appears to be silent to the outside world – a system that sounds like science fiction but appears entirely possible.The Guardian