Scientists at the South Federal University's Neurotechnology Research Centre are investigating the possibility of creating a brain-computer neural interface based on brain activity generated by the mental utterance of word commands.
The first results of these studies have been published in the journal Biomedical Signal Processing and Control.
Such a neural interface could benefit those who have lost real speech as a result of stroke or other pathology, such as tetraplegia or paralysis of the facial muscles and chewing muscles, but who are still conscious and able to think and feel, and who find themselves in social isolation. A person who has lost their speech has difficulties in interacting with the world around them and the people around them, which they are sometimes unable to overcome. SFU scientists are looking for a way to solve this problem by creating channels of communication with the world based on brain activity recorded as an electroencephalogram (EEG) and interpreted using artificial neural networks and artificial intelligence technologies.
"Speech is a complex cognitive process, the realisation of which requires the coordination of a number of large hemisphere cortical structures, both in real and internal speech. We investigated the EEG by applying neural network classification techniques to recognize patterns of brain electrical activity specific to mentally spoken words. The recognition accuracy of the patterns of brain activity corresponding to rest and mental speech was over 90%, and that of individual mentally uttered words more than 50% at a random selection level of 10%," said Dr Oleg Bakhtin, a leading researcher at the Research Technological Centre for Neurotechnology, Candidate of Biological Sciences.
According to the scientist, similar research results are currently associated with the registration of so-called electrocorticographic activity by microelectrode matrices implanted directly into the brain.
"Our approach could enable a brain-computer interface without the use of surgical procedures. It also has the advantage of using internal speech, which is natural for humans, for communication, as well as efficient neural network algorithms, which are being developed by the specialists of the Centre for a long time. It is obvious that the EEG method we use to register the activity of the brain areas related to speech has obvious advantages over methods based on surgical intervention in its activity to implant microelectrode arrays," added Oleg Bakhtin.
In the future, scientists at Southern Federal University plan to continue experimental research for a more detailed study of EEG indicators of mental speech, taking into account its phonological and phonetic features.