2019-03-01 07:56:48

As reported by [gizmodo]:

A team of scientists has been able to translate brainwaves directly from the audiotory cortex of the brain and translate them into speech. To do this they selected a number of epilepsy patients undergoing brain surgery, and invasively inserted electodes directly into their auditory cortex, they then had them listen to people count from zero to nine and recorded their subjects EEG readings, then used the resulting data to train a Deep Neural Network to translate them into speech. The hope is that they will one day be able to create vocal prosthetics for those who are no longer able to speak by being able to translate brainwave output directly into sound, as opposed to stringing together pre-recorded sounds as some Voice Synths do.

You can listen to the reconstructed brainwave audio [here].

-BrushTone v1.3.3: Accessible Paint Tool
-AudiMesh3D v1.0.0: Accessible 3D Model Viewer

2019-03-01 08:07:55

whoa

Facts with Tom MacDonald, Adam Calhoun, and Dax
End racism
End division
Become united

2019-03-01 08:47:21

That's actually really neat, wonder where it's gonna go? Although slightly disappointed they only gave us samples from 1 of the people, unless I'm missing something. Would've been interesting if we could compare.

"You know nothing of death... allow me to teach you!" Dreadlich Tamsin
Download the latest version of my Bokura no Daibouken 3 guide here.

2019-03-01 13:35:58

Fascinating. Could this lead to wireless monitoring of a person's thoughts? And is this the first step toward the sci-fi idea of being linked to a network via implants in the brain? It's interesting and a little bit creepy too.

2019-03-01 14:44:46

OK, seriously, this is awesome and I wonder where this will go. The fourth sample was amazing.
On a more observational note and not so serious, The second sample sounds kind of creepy, but that's nothing compared to the third, that one sounds like some sort of science fiction villain, I had to pause it halfway through and compose myself and continue playing!

2019-03-01 23:25:43

@3
I'm not sure how they processed the data, whether it was collectively or individually, though the results they posted consist of 30 minutes of training data. Auditory information is also processed differently from person to person, I think they would have been interested in getting a larger data set to see what the average of that might have been, but only had so many subjects to work with.

@4
There have been people who've had analog jacks implanted in their brains to move cursors around on a screen, so its a bit of an evolving field. These current experiments though make use of Neural Networks to help decode brainwaves, there have been some amazing breakthroughs with portable [MEG Brainwave Scanners] using quantum sensors and Neural Networks to decode brainwaves, they kind of look like Phantom of the Opera masks. This current experiment though doesn't process "thought speech" or attempts for the subjects to speak, but rather the sounds they hear. But, theoretically it could in the future allow others to listen to peoples thoughts in an entirely unsettling way.

-BrushTone v1.3.3: Accessible Paint Tool
-AudiMesh3D v1.0.0: Accessible 3D Model Viewer