In the XXI century, the technology of subvocalization is experiencing a revival – a group of scientists from MIT under the leadership of Arnav Kapur is developing a headset for computer control without the participation of voice. They managed to almost completely exclude the acoustic component when transferring words from person to machine and vice versa. This is not a superstructure over existing systems, but a principled new interface called AlterEgo.
Unlike the laryngophone, this headset reacts not to mechanical vibrations of the skin when pronouncing words, but to electrical activity in the muscles and nerves of the larynx, jaws and tongue. The user does not need to literally open his mouth, let in air and say the words, enough “clearly and clearly” imagine how he pronounces the team. A set of electrodes counts the impulses, the neural network matches their pattern to the database and decodes what the person said.
During the research, Kapoor's team reduced the number of sensors from one and a half dozen to just four, determined the optimal locations for their location, developed algorithms for analyzing subvocalization data. The response from the computer is also transmitted silently, through the vibrations of the bones of the skull. Therefore, the experimental headset looks “broken” – it does not have a familiar microphone and earpiece. And the process of dialogue is more like reading thoughts, which can not be overheard from outside.
The vocabulary of the system is still symbolic, but it is perfectly suitable for the transmission of short codes, for example, the designation of chess moves. After 15 minutes of personal customization, volunteers within 90 minutes of silent communication allow only 8% of errors. This is so encouraging to Kapoor's team that scientists are already thinking about creating a new human-machine interface. Why do we need to open the skull to the user and implant some transmitters to read the thoughts, if there is an opportunity to go another, simpler and painless way?js.src = “&version=v2.8”; 'script', 'facebook-jssdk'));