Slashdot Mirror

← Back to Stories (view on slashdot.org)

Major Breakthrough in Direct Neural Interface

Posted by Zonk on from the ready-for-my-dni-thanks dept.

jd writes "In a major breakthrough, neurologists are reporting that they can decypher neurological impulses into speech with an 80% accuracy. A paralyzed man who is incapable of speech has electrodes implanted in his brain which detect the electrical pulses in the brain relating to speech. These signals are then fed into computers which covert these pulses into signals suitable for speech synthesis. As a biotech marvel, this is astonishing. Depending on the rate of development it is possible to imagine Professor Hawking migrating to this, as it would be immune to any further loss of body movement and would vastly accelerate his ability to talk. On the flip-side, direct brain I/O is also a major step towards William Gibson's Neuromancer and other cyberpunk dark futures."

7 of 308 comments (clear)

  1. More info by niceone · · Score: 4, Informative
    The BBC article is pretty light on detail, and the New Scientist one is subscribers only, but there is more stuff here.

    They have hooked up to 41 neurons and:

    For now, the team is focusing on the building blocks of words. In a series of experiments over the last few years, Ramsey has imagined saying three vowel sounds: "oh", "ee" and "oo". By watching his brain activity, the researchers have been able to identify distinct patterns associated with the different sounds. Although the data is still being analysed, they believe that they can correctly identify the sound Ramsey is imagining around 80 per cent of the time
    --
    ccalam - acoustic versions of new songs.
  2. Re:Really accurate? by rucs_hack · · Score: 4, Informative

    I'm guessing the 80% comes from the fact that this is an issue of the linear separability of signals. Its generally hard to get reliable sensitivity/specificity measures over this that anyone is going to take seriously.

    Sensitivity = percentage number of correct identifications
    Specificity = corresponding percentage of incorrect identifications at each measured sensitivity.

    Probably they can get up to 90%, but from experience I would say the rate of false positives at this sensitivity likely is moving towards exponential increase. It's better to stop at 80%, at least when something is in the early stages.

    This is just guessing of course, I have no understanding of their research, but going from my own work on non linearly separable sets, I'd say this is what's happening.

  3. can still communicate by circletimessquare · · Score: 4, Informative

    http://en.wikipedia.org/wiki/The_Diving_Bell_and_the_Butterfly

    he could blink. that's it. yes or no. and with that ability, letter by letter, he wrote a book (with the help of some very patient nurses/ assistants)

    it's coming out as a movie soon too i think

    --
    intellectual property law is philosophically incoherent. it is your moral duty to ignore it or sabotage it
  4. Re:What? by klenwell · · Score: 5, Informative

    I believe Antonio Damasio addresses this question in one of his books. Apparently, a fortunate side-effect of this condition is it impairs the part of your brain that would normally find this horrific and intolerable and leaves you with a weird sense of acceptance and well-being (IIRC). Otherwise, I guess you just blink a lot and hope they keep the feeding tube hooked up.

    --
    Innovation makes enemies of all those who prospered under the old regime... -- Machiavelli
  5. 80% accuracy... by uwbbjai · · Score: 3, Informative

    It reads: "Dear Aunt, let's set so double the killer delete select all"

    What do you want to decipher today?

  6. Research posters by FleaPlus · · Score: 4, Informative

    For those curious, this speech prosthesis research was presented in a number of posters at the Society for Neuroscience (SfN) conference a couple weeks ago. Their six SfN posters can be found on their website here, covering topics like the circuitry they developed, Bayesian signal analysis, and so forth:

    http://migrate.speechprosthesis.org/DNN2/SpeechProsthesisHome/tabid/52/Default.aspx

    There's also a nice blog entry on this over at Neurophilosophy:

    http://scienceblogs.com/neurophilosophy/2007/11/speech_prosthesis.php

  7. Vinge, not Gibson by wurp · · Score: 3, Informative

    Gibson didn't invent cyberspace. Vernor Vinge invented cyberspace (although I don't think he coined the term) in True Names.

    If you haven't read it, I highly recommend it. Read True Names to get a notion of the profound visionary Vernor Vinge is. (Remember it was published in 1981).

    Then read Rainbows End with your newfound respect for Vinge's powers of prognostication, and recognize that you're seeing into the near future.