dcsimg
Results 1 to 4 of 4

Thread: Analyzing raw sound data and using AI to make predictions

  1. #1

    Thread Starter
    New Member
    Join Date
    Feb 2016
    Posts
    8

    Analyzing raw sound data and using AI to make predictions

    I wanted to build a simulation, and effectively teach a computer to enjoy music, and tell me when it liked a song.

    Humans find pleasure in symmetry and predictable (but not verbatim) patterns in music. When we "partially" predicted the melody of a song we heard for the first time, it's more enjoyable when we hear it.

    Why is symmetry and predictability pleasurable? Because they're easy to process.

    What if a machine could make educated guesses on the next X bits of sound data based on the previous X bits, and then reward itself with "enjoyment" that would scale using an algorithm that took into account:

    1. How close it's estimate was to the actual sample
    2. The CPU/memory/system resource usage needed to estimate

    Obviously like a human, the lower resources needed to "perceive", the better it was.

    I noticed that .WAV files tend to display a graphical soundwave (think the visual representation of the audio you see in recording programs like Audacity on the timeline). Would it make more sense to do the analysis I mentioned visually (i.e. generate a simulated "guess" .WAV graphically, and compare it graphically with the actual visible sound bit)? Or would comparing the raw data make more sense?

    How possible/impossible is this?

    Is VB.NET the wrong language for this?

  2. #2
    Frenzied Member
    Join Date
    May 2014
    Location
    Central Europe
    Posts
    1,285

    Re: Analyzing raw sound data and using AI to make predictions

    Or would comparing the raw data make more sense?
    the graphic you see is only the Amplitude, you cant "see" the frequency in this Picture, so you'd Need to work on the raw data.

    How possible/impossible is this?
    definitely a very complex and ambitious task

    Is VB.NET the wrong language for this?
    no, it's as good as the others i'd say.

    from how you describe the Problem and your questions, i'd think you have a very incomplete Picture of the Topic and underestimate the complexity of your intend. i'd suggest you read up on what pcm wave data is, how it produces Sound, what an FFT is and how the same data is expressed in the frequency Domain, then consider how many Bytes of data you Need to represent one second of Music in time and frequency Domain and then you might see that it is not a Task of comparing some Bytes here and there but doing complex Analysis on quite big amounts of data. that will help you to correctly assess your plans.

  3. #3
    Fanatic Member namrekka's Avatar
    Join Date
    Feb 2005
    Location
    Netherlands
    Posts
    639

    Re: Analyzing raw sound data and using AI to make predictions

    Quote Originally Posted by logiconly View Post
    ....Humans find pleasure in symmetry and predictable (but not verbatim) patterns in music. When we "partially" predicted the melody of a song we heard for the first time, it's more enjoyable when we hear it.

    Why is symmetry and predictability pleasurable? Because they're easy to process.....
    Do we like out of tune?
    So how is a melody build up?

    .....Obviously like a human, the lower resources needed to "perceive", the better it was.....
    Is it???

  4. #4
    Frenzied Member
    Join Date
    May 2014
    Location
    Central Europe
    Posts
    1,285

    Re: Analyzing raw sound data and using AI to make predictions

    Quote Originally Posted by namrekka View Post
    Do we like out of tune?
    So how is a melody build up?



    Is it???
    yeah, these are additional problems, i forgot

    and why do some people like songs that others hate? do you find alien scales like chinese ones appealing?

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  



Featured


Click Here to Expand Forum to Full Width