Apr 20, 2017 11:43 AM EDT
One Japanese research collaborative team has explored a new kind of brain technology. A subject, with electroencephalogram (EEG) readings, can utter numbers zero to nine with 90 percent accuracy. It can also identify 18 kinds of Japanese monosyllables from EEG signals with 60 percent accuracy. Soon, there might even be an EEG-activated typewriter.
This brain technology research breakthrough will be presented at Interspeech 2017, August, in Stockholm. The goal of the conference is to "provide a broad approach to speech communication issues integrating speech technology and pragmatic aspects of human conversational speech communication behavior in different interaction contexts."
The brain technology team gathered EEG data of subjects who spoke Japanese digits as well as monosyllables. The group then undertook experiments that showed digit and monosyllable recognition. So far, speech decoding through EEG signals has not been made possible. There has not been enough information regarding the use of powerful algorithms that are based on knowledge or related kinds of machine learning, according to Phys.org.
With the help of a smaller training dataset, the team created a different kind of research framework for higher achievement in brain technology. It was a novel kind of framework that was based on holistic pattern recognition. It made use of category theory or composite mapping. In this, a dual space as well as a tensor space, with exterior algebra was employed.
The experiment showed that spoken digit recognition could be achieved from EEG signals. It also gained 90 percent recognition accuracy. However, there was also the attainment of 61 percent accuracy in 18 Japanese monosyllable recognition. This was quite an achievement that outperformed earlier research on brain technology.
Hence, a brain-computer that can identify unvoiced speech or imagery is aimed to be developed by Emeritus Professor Nitta and his group. It is a brain technology that might restore speech to those who have lost the ability of voice communication. The technology is expected to provide healthy persons with a natural interface and no limitations. The team also hopes to create a device that can be operated with fewer electrodes and also connected to smartphones in a few years.