An international group consisting of American and German researchers have succeeded in decoding signals generated in the human brain in natural continuous speech. And the details of the transcript were used to convert the "mental" speech in its text representation with a rather high degree of accuracy. Further development of this technology may be the first step towards the creation of systems makeupalley computers and other equipment, new technologies of communication between people, and more.
A system of "Brain-to-Text" to produce entry signals that can be read by means of a matrix ECoG electrodes (electrocorticographic) located in the appropriate places opposite the frontal and side areas of the cerebral cortex. In the experiments, was involved in seven volunteers who were in one of the medical institutions where they took a course of clinical treatment available to them in nervous disorders, including epilepsy.
All the participants read out loud the standard text, which consisted of a limited set of certain words. The collected data of brain signals have been "fed" to the computer, which by means of algorithms of computer investigation and analysis extracted the most probable sequence of signals corresponding to particular words. After that, by using a fairly simple algorithm, the data was converted to text format.
Unfortunately, the accuracy of such a system still leaves much to be desired. Under normal conditions, the number of errors in the determination of the spoken human words was at the level of 25 percent. And in case if the patient for some reason was unable to pronounce the words absolutely right, changing some of the phonemes in the pronunciation, the number of errors was increased to 50 percent.
But scientists are not going to stop on achieved results. They explain the high percentage of errors that the analysis of data was performed using universal software, not specialized, which should work much more effectively. When this highly specialized software will be created, the percentage of committed system errors are drastically reduced and these systems will become the "bridge" that will allow you to communicate with people, unable to speak due to various reasons.
It should be noted that these works were commissioned by the research Department of the U.S. army, supported by several research funds. Recording brain signals was carried out by specialists of the Medical center Albany (Albany Medical Center) Albany, new York, and the algorithms of signal processing and automatic speech recognition have been developed in the laboratory, Cognitive Systems Lab, Karlsruhe Institute of Technology (Karlsruhe Institute of Technology KIT), Germany. published
P. S. And remember, only by changing their consumption — together we change the world! ©