In recent days, scientists have made some very encouraging progress in regard to brain-to-computer interfaces. They have come up with machines that can transform thoughts into controls for wheelchairs, drones and music. Now Japanese-based researchers report that they have gone further and developed new technology that is able to identify Japanese words and also deduce the single-digit number that is on the mind of a subject with an accuracy of 90 percent just by scanning their brainwaves.
The system was designed by researchers at the Toyohashi University of Technology of Japan and it utilizes an EEG(electroencephalogram) cap. These are headsets that use electrodes that are positioned on the scalp to scan electrical signals emanating from the brain. Even though they can be applied for a wide-range of uses, one area in which they hold special promise is in the improvement of the lives of people with disabilities.
For instance, in 2010 we witnessed a home-based EEG cap setup that allowed people suffering from locked-in syndrome(characterized by the brain being active while the body is not) to communicate by choosing characters on a virtual keyboard just by concentrating on them. During the same year, researchers based at the University of Utah showcased an implantable version (one that does note penetrate the brain) that can translate brain signals into a definite number of words. During the study, the team of researchers put grids of miniscule microelectrodes onto the brain speech centers of a volunteer who was suffering from serious epileptic seizures. These non-penetrative microelectrodes, also known as microEC0Gs, are implanted below the skull but sit atop the brain without protruding into it. Since they do not penetrate brain matter, they are thought to be safe enough to be placed on the brain’s speech areas. The same cannot be done with the penetrative electrodes that are normally used in experimental devices to assist paralyzed persons to control an artificial arm or a computer cursor.
The Japanese researchers also harbor hopes of one day enabling people with speech disability to communicate. They state that transforming EEG signals into words has in the past been restricted by the volume of data that these systems can compile. Now they are using a new signal interpretation method that they claim is based on all-inclusive pattern recognition coupled with machine learning. This enables it to attain high performance using only a small data set.
The system is highly effective, to the extent that it is able to recognize spoken numbers between zero and nine with an accuracy level of 90 percent. What is probably more encouraging is that the system was able to identify 18 Japanese words (single-syllable) with an accuracy level of 61 percent, which enhances the possibility of developing a brain-activated typewriter.
With more development, the team envisions that its system will assist people with disabilites and also facilitate a smoother computer interface for healthy individuals. It states that it intends to fine-tune the technology and design a device that can be plugged into smartphones in the next five years. It will provide updates on its progress during the Interspeech Conference that is scheduled to be held in Sweden during the month of August this year.