University of Columbia Researchers Translate Brain Signals Directly Into Speech (columbia.edu)

dryriver writes: There is good news for people who have limited or no ability to speak, due to having suffered a stroke for example. Researchers at Columbia University have managed to turn brain signals in the auditory cortex of test subjects into somewhat intelligible speech using a vocoder-like system with audio output cleaned up by neural networks. The findings have been published in the journal Nature. Here's an excerpt from the Zuckerman Institute's press release, which contains example audio of a number sequence being turned into robotic speech: "In a scientific first, Columbia neuroengineers have created a system that translates thought into intelligible, recognizable speech. By monitoring someone's brain activity, the technology can reconstruct the words a person hears with unprecedented clarity. This breakthrough, which harnesses the power of speech synthesizers and artificial intelligence, could lead to new ways for computers to communicate directly with the brain. It also lays the groundwork for helping people who cannot speak, such as those living with amyotrophic lateral sclerosis (ALS) or recovering from stroke, regain their ability to communicate with the outside world."

Comments

Brain microphone, not brain speech synthetizer

[abies]

Please note that while it seems like they cracked translating thought into speed, it is really about something very different. They decode what person is hearing into sound. This is very, very far from being able to interpret actual thoughts - they attach themselves to very early stage of hearing process, where things are still looking very much like sound waves. Additionally, it is tested on just 10 words (numbers from 0 to 9). On top of that, given neural nets are magic, I wouldn't be surprised if instead on brain activity, they focused on some measurement artifact, which is caused by skull vibrating to the sound, turning all that setup into overcomplicated and underperforming microphone. Sure, it is a worthy research, but it is light years away from thoughts to speech conversion.

Re:Brain microphone, not brain speech synthetizer

[teslar]

Yes, this.

The only thing I would add is that we do know that higher-level cognitive processes re-activate sensorimotor parts of the brain; so thinking the number "nine" might cause some activation similar to hearing that number.

So, assuming they didn't indeed just pick up on some non-neural artifact, this shows that there is potential - at least in the ideal case where we know the signal is as good as it's ever going to get. It's one step in the direction of being able to deal with weaker, noisier versions you might get from reactivation.

But yeah, noone's going to read your thoughts anytime soon.

Re: "I'm hungry. Boobs! What should I order? Boobs

[Vintermann]

It would be creepy as hell, but very cool, to get a recording of the voices schizophrenics hear.