The new method of voice synthesis has lost its delay and voices the user's thoughts in real time.
The new brain implant has allowed a woman who has lost the ability to speak, to turn thoughts into real -time language, according to Science Alert.
An implant in a woman's head analyzes the brain activity at an interval of 80 milliseconds and instantly turns it into a synthesized version of voice.
The main difference from old methods is the lack of delay - most existing technologies require the processing of the whole text before its voice, which causes pauses between the formation of thought and its vocalization. This made communication unnatural and caused discomfort from users. In addition, the previous methods required users to teach the system, mimic movements during sounding, which could complicate their use for persons with serious speech disorders.
To solve these problems, researchers from the United States have trained a neural network based on the sensorimotor activity of a 47-year-old brain cortex. She "spoke" 100 unique sentences in silence, using more than 1000 different words.
The voice of a woman was synthesized on the basis of records of her previous speech, which made the result more natural and close to her own voice.
The research group also tested the efficiency of the system based on 50 phrases from a smaller set of words. In both cases, the average rate of translated words per minute was almost twice as high as in previous analogues.
It was also noted that the system can interpret neural signals that represent words to which it has not been trained, which opens up opportunities for further improvement of technology.
The study is published in the Nature Neuroscience .
