Decode conversations in real time through brain activity

, published in the journal Nature Communications, describes a neurodecoder that converts brain activity associated with question-and-answer conversations into written records.
the cerebral cortical layer contains different regions in which neural activity encodes speech perception and generation. Studies have shown that this brain activity can be decoded, but previous studies have focused on decoding hearing tasks separately.
Edward Chang of the University of California, San Francisco, and colleagues decoded speech perception and generation through brain activity in a simulated question-and-answer dialogue experiment. The data was used to train voice detection and decoding models after three patients who were being treated for epilepsy were asked a series of questions and answered verbally through a set of prescribed answers. The subjects then listened to a series of questions and answered loudly with the answers of their choice.
using only the neural signals recorded in the conversation, the authors were able to detect when the subjects were listening, when they were ing2, and were able to predict what they were listening to or saying. By decoding questions, they were able to use the information they received to improve the accuracy of the decoded answers (because some of the answers apply only to specific questions), and increased the decoding accuracy of generating and perceptive speech to 61% and 76%, respectively.
further work is needed to decode imaginary speech responses so that individuals who are unable to speak due to injury or neurodegenerative diseases can use the technology. (Source: China Science Daily, Tang 1 dust compilation)
Related paper information: