These AI-booted brain-machine interfaces give voice to two women

Losing speech as a result of disease or brain damage often leads to isolation for people with this condition. In order for them to be able to express their thoughts, emotions and needs again, scientists have long been working on brain implants combined with powerful AIs. Two recent studies have given the voice back to two women. One is paralyzed following a stroke, while the other suffers from a progressive neurodegenerative disease.

Thus, in the United States, with the solution of the researchers of theuc san francisco (UCSF) and UC Berkeley, the woman who suffered a stroke was able to speak out via a virtual avatar. The AI ​​algorithms used transform brain signals into speech and facial expressions. For this, the scientists implanted a thin rectangle of 253 electrodes on the surface of his brain on the critical areas related to speech. These are the same areas that would have animated the muscles of the lips, tongue, jaw and voice box without this stroke. These electrodes are connected by a cable to computers. It took the patient working with researchers to train the AI ​​algorithms to recognize her unique brain signals.

Thus, for weeks, she had to repeat different sentences with a base of 1,024 words. For the AI, it wasn’t about recognizing whole words, but rather about determining them with respect to less than 39 phonemes. It was therefore necessary to design a voice synthesis system made up of old audio recordings of the patient’s voice. The process is capable of decoding a large vocabulary and transforming it into text at a speed of 78 words per minute, with an error rate of 25%. This is almost half the natural conversation speed, but it is already a huge improvement for this woman who can now communicate with her husband. Researchers are now working on creating a wireless version of this brain-machine interface.

Ann suffered a stroke that left her speechless. Electrodes were implanted in his brain, then connected to the computer which, using an AI, translates the movement of his lips and his attempt to speak into words spoken via an avatar. © Noé Berger, UCSF

Predict words from phonemes

In the other case, the laboratory of Stanford Medicine, managed to transcribe in the form of texts, the cerebral activity of a woman suffering from a serious degenerative neurological disease. This 68-year-old patient remains able to formulate instructions to generate phonemes. To exploit them, the Stanford Medicine researchers implanted two tiny networks of sensors on the surface of the patient’s brain. They are found in two distinct regions of speech production. Each of these networks has 64 electrodes. They connect to the cerebral cortex with a depth of 3.5 millimeters.

Again, the AI ​​was trained to distinguish nuances of brain activity emitted around the wording of 39 phonemes. After 25 sessions during which 260 to 480 sentences were repeated, the system was able to reconstruct the words associated with these phonemes. The error rate was limited to 9.1% over fifty words. This rate increased to 23.8% for 125,000 words. The conversion speed in this case was 62 words per minute.

In these two situations, if the processes are promising, they remain, for the moment, unfortunately limited to the environment of a laboratory, but it is already a huge progress!