Brain Waves: Neuroprosthesis with a Large Vocabulary

Brain Waves: Neuroprosthesis with a Large Vocabulary

Neuroprostheses translate nerve signals directly from the brain, for example to control prostheses of the hand. The power of thought can also be used to move the cursor on the screen or to move a wheelchair in space. In the future, the interface should help paralyzed people again communicate directly with their environment. Because in many of those affected, the ability to speak is fully intact, for example as– either off-patient.

Over the years, intensive research has therefore been carried out into the interfaces that are supposed to “understand” the fictional language. Among other things, it has already become possible to conduct simple conversations between At least two people to translate in many cases, Such experiments are usually done with epileptic patients who have electrodes placed in their brains because of their disease. The activity of some nerve cells can be read directly there. even non-invasive methods, such as electroencephalography (EEG), is now being tested.

electrodes in the brain

Other studies work directly with affected people, such as one published in the journal “Nature Communications.” Work team’s edward chang From the University of California, San Francisco: The subject was a 36-year-old patient who was almost completely paralyzed after a severe stroke and could no longer speak. With the help of only small movements of the head, he operates a computer-aided interface.

For the experiments, a credit card-sized implant with 128 electrodes was inserted into a young, cognitively unaffected individual. Researchers already had their first breakthrough last year in the New England Journal of Medicine informed of. The program used at the time was developed with the help of approximately 50 training units. “reading or learning by meditating” trained while the patient attempted to articulate one of the 50 given words.

See also  The glider-helicopter that flew on Mars now has the hardware of a 7-year-old smartphone

In addition to machine learning methods from artificial intelligence, classic natural language models were also used, which, for example, indicate the probability of occurrence of words in a specific context. In this way, the program actually learned to convert brainwaves into language. Finally, it was also possible to recognize entire sentences from the limited vocabulary. However, the error rate was a quarter.

mentally speaking

The approach tested for the present study should not only be more accurate and comprehensive, but also more direct. This means that the test taker no longer had to try mentally – that is, even physically – to speak, but was able to think of the word or letter directly. This makes the process less unnatural and is therefore perhaps slightly more suitable for everyday use for those affected, write Chang and Co.

Incidentally, other research groups are also trying to make the brain more readable by circling through the motor system, for example in the form of “mind writing” – as opposed to purely fiction writing by hand, such as A team just reported last year, As the new work now suggests, it may not be so important to computers whether the brain motors or mentally produces words.

few letters, many words

Chang’s team found a fairly simple trick to expand the terminology of neuroprosthesis without much effort. Instead of thinking up whole words, this time the subject was to write them down. The study’s authors emphasize that this would allow a much larger number of words to be represented with only 26 letters. In order for the machine to better “understand” the letters, the International Alphabet Table (“Alpha, Bravo, Charlie, …”) was used. For training and testing, the classic models of artificial intelligence and language processing were reimagined.

See also  iOS 16 brings back battery percentage, users are confused

In fact, in the vast majority of cases, the computer was able to recognize mental sentences generated by the letter method from a basic vocabulary of 1,152 words. The average error rate per character was just over six percent, and about 30 characters could be processed per minute. This glossary will be of great help for everyday communication. Simulation calculations by the researchers show that the vocabulary can be increased to 9,000 words without any major loss. Even so, the average error rate would be just a little over eight percent.


Please enter your comment!
Please enter your name here