Science has taken a step further in understanding the human brain: researchers were able to decode brain waves related to speech and translate them into words.
The machine’s algorithms were programmed to translate brain waves in real-time to form words and sentences, with an error rate of just 3%. Other similar attempts had already been successful in brain-machine interaction, but always in a limited way, managing to capture only fragments of words.
The new technology was tested on 4 volunteers reading 30 to 50 sentences aloud with electrodes monitoring their brain activity. The data was subsequently translated by a computer in the form of written words.
The translation of brain waves into words occurred in two steps: first, the machine analyzed the pattern of waves generated after repetitions of vowels and consonants. Then, another system joined each of these fragments to form the words.
During the decoder training, the machine learned phrases that were not in the final test. And since the technology is managing to identify words and sentences, it may also be able to decode sentences for which it has not been trained.
“Although we should like the decoder to learn and exploit the regularities of the language, it remains to show how many data would be required to expand from our tiny languages to a more general form of English,” wrote the University of California researchers in an article published in Nature Neuroscience.