Columbia University researchers have developed a system that can produce clear, intelligible synthetic speech through computer processing of human brain activity.
The researchers first analyzed the brain activity of volunteers as they listened to other people's voices, which generated electrical patterns in the brain's acoustic cortex that mirror those from speaking, but are easier to decode. They then used an artificial intelligence program, called a vocoder, to process brain signals from patients listening to a series of numbers and use them to synthesize speech.
These sounds were analyzed further and cleaned up with a neural network, resulting in a series of words.
Said Columbia's Nima Mesgarani, "If the wearer thinks 'I need a glass of water,' our system could take the brain signals generated by that thought, and turn them into synthesized, verbal speech."
From Financial Times
View Full Article - May Require Paid Subscription
Abstracts Copyright © 2019 SmithBucklin, Washington, DC, USA