New System Generates Speech from Brain Physiology

Join the discussion
Registration is free. Ask a follow-up in this thread, or start your own.
1 reply · 2K views
Messages
2,831
Reaction score
12,430
TL;DR
Scientists published a study describe using information from brain recordings to generate understandable speech.
Scientists described using information from brain recordings to generate understandable speech in a proof of principle study.
Invasive recordings of brain activity normally involved in the production of speech sounds, at the level of controlling muscle movement, have been associated with the speech sounds produced. This has allowed recordings of brain activity to drive the generation of understandable machine generated speech. The invasive nature of the electrode placement (inside the skull) makes finding trial subjects more difficult.
In the long run, this process may be developed to being clinically useful, but it is not ready yet.
Medical problems affecting the speech motor areas (where the recordings are made) could rule its use out in particular cases.
Here is a NY Times article on it.
Here is the original article in Nature which is behind a paywall.
 
  • Like
  • Love
Likes   Reactions: DennisN, atyy and berkeman
Biology news on Phys.org
BillTre said:
Summary: Scientists published a study describe using information from brain recordings to generate understandable speech.
When I first read the summary I found it hard to believe, but then I read the abstract,

Abstract said:
Recurrent neural networks first decoded directly recorded cortical activity into representations of articulatory movement, and then transformed these representations into speech acoustics.

which I suppose means that the system has to be "taught" depending on which person is using it, which makes the technology understandable and feasible to me.

And further down the abstract reads

Abstract said:
Decoded articulatory representations were highly conserved across speakers, enabling a component of the decoder to be transferrable across participants.

which I find very interesting and a bit surprising. And this makes me wonder if and how different the conserved decoded articulatory representations would be for different languages, e.g. English and, let's say Spanish.

Anyway, this is amazing and inspiring research! Thanks for posting!
 
  • Like
Likes   Reactions: atyy and BillTre