TED Conversations

Jari Hiltunen

unemployed and looking, AB Enzymes

This conversation is closed.

Damaged brains and voice recognition

I met yesterday person who have brain damage due to bleeding in the brain. He had lost his ability to speak properly, meaning of words was mixed like "bird" means "light" etc.

He has also some words/speech sounds which I do not recognize. I presume this may be typical condition with injured brains.

I wondered, could voice recognition software, like Google's voice recognition be adjusted so that person may "teach" application to translate voices? Meaning, if person says "bird", Google (or whatever app) translates it to "light", or maybe just some speech sounds to meanings (educate injured person to pick meanings for sounds).

Have you seen such applications anywhere? Shall be rather easy to implement.

Share:
  • thumb
    Mar 22 2014: Hi Jari, There is lots of research happening in this field. My son, Matt Winn, is a PhD at the University of Wisconsin and he is focused in the improvement of speech recognition. Here is a link on his initial ideas for this research.....

    http://www.mattwinn.com

    and here is the link for further information on this subject.

    http://www.waisman.wisc.edu/bhl/about_people/People15.html

    Of course, each case has its own challenges and when the brain is damaged, you really need the best of the best in this field to get some answers. Communication is so important and I wish you the best in finding some help.
  • thumb
    Mar 6 2014: somewhat relevant: http://en.wikipedia.org/wiki/Aphasia

    I don't know which kind of aphasia that person has, but it must be something like that.

    If the meaning of words is scrambled, then it would be a case-by-case way to "translate" it (as people with similar speech disorders would probably have other types of scrambling).

    Software should thus be that it can learn the specific changes. That would mean the program would need quite some training (with the patient and an assistant).
    • Mar 18 2014: I agree with Christophe Cop,

      If the case were the only change is bird to light and light to bird, it wouldnt be a dificult adaptation on a software that alread exists.

      But its complex to do learning machine, and introduce new "software cenarios", were it have to reconise a assistent voice, or acept other kind of inputs like pictures choosing or commands to teach the machine, turn to learn mode or function mode

      And we would probably have to insert a lot of "commom problems" solve functionality based on studies and probably would be a lot missing.

      I think there is no software today ready to use for this but i believe it is a great project
      That would to help people, study disorders and get involved or even develop great software tecniques.
  • Mar 6 2014: Hi Dear Jari,I think it isn't easy to do,because everyone's brain is unique,so the damage of the brain is individual too.And it is cocern about brain.we people are still been confused a lot how we people learn things,how our brains reacting the learning material,how we people have creative ideas?all are still sort of mysterious or on the way to discover.

    But I am very postive to look forward to the creation comes to those people they needed,only time thing.
    • thumb
      Mar 6 2014: I understand. It must be individually adapting application for persons who have lost their ability to speak later age, when you have already context and meanings of words in your brains.

      As an example, if I show picture of a bird, and ask disabled person to say "bird", whatever sound person is able to say for "a bird", then meaning is bird. I guess that with some 1000 words you can already have pretty good conversations in any language, why not in "disabled language" as well.

      I am surprised how well Google's voice recognition in Android phone works, and in many languages (like finnish). I don't know what kind of algorithms it is using, but they must be built on some sort of voice samples. If those voice samples can be individually adjusted, then I guess it is possible to make an application, which translates recognized disabled speech to understandable speech.
  • thumb
    Mar 6 2014: I don't think reinforcing incorrect responses is the best direction to go, but keep looking the answers will come.