Pattern-matching with analyzer~

    Oct 10 2011 | 8:25 pm
    I'm looking for a pattern-matching object or system that can take the continuous output of something like analyzer~ and notice time based trends. Specifically the trends I'm seeking to identify is the sounds of a siren passing by on a street (long slow change in pitch, quick "woop" at the right frequency followed by certain amount of silence, etc).
    Anyone have any recommendations or ideas on how to approach this? Might the FTM gesture follower v( be a candidate?
    I imagine that this will also involve some serious noise reduction and filtering as well.

    • Oct 13 2011 | 2:33 pm
      Any ideas guys?
    • Oct 13 2011 | 2:52 pm
      I have nothing to add here but my interest as well.
    • Oct 13 2011 | 7:41 pm
      Just off the top of my head, if you can identify exact strings that are produced , OR can map the desired output into a reproducable set of strings, the REGEXP world can be your pattern matching friend... just an idea. & yes, the challenge is identifying the output you want to match on, and getting the desired input to always produce the 'matchable' strings... detect ranges of input,important 'cue' values, etc, etc.
    • Oct 18 2011 | 3:13 am
      I'll look into regexp! Thanks for the recommendations. Any ideas for noise reduction/filtering?
    • Oct 18 2011 | 9:27 pm
      depends what you are looking for: a noise gate with adjustable threshold would be a strong need, to avoid the analysis from trying to analyze noise, etc.: you need a strong enough signal to analyze!! Then, avoid the extremes of low and high freq... low is tough to analyze, high has lots of randomness, also tough to get clean stuff from... just a off the cuff response... g'luck!!