Onset detection for slow attacks - looking for advice

    Apr 18 2013 | 10:46 am
    I'm working on piece for hexaphonic guitar using ebow and need to do some onset detection. The audio from the individual strings is basically monophonic but noisy and with some crosstalk from adjacent strings.
    The attacks can be very slow.
    I've looked at few approaches but so far the detection is rather late and some events don't get detected at all.
    Any Ideas?

    • Apr 18 2013 | 11:01 am
      Hi Richard, if you don't get any success here, it might help to get in touch with my colleague Ricky Graham, Assistant Prof of Guitar Technology at Stephens University, NJ:
      He's a good friend of long-standing and if you mention my name, he'll be very pro-active and helpful. He's done considerable work in the area of hexaphonic string output and envelope following (albeit using PD on the software side).
    • Apr 19 2013 | 8:34 am
      Hi Brendan,
      Thank you for that. I will get in touch with him.
    • Apr 19 2013 | 11:24 am
      >He's done considerable work in the area of hexaphonic string output and envelope following
      wow, that is a hardcore specific reference right there! :) Love it.
    • Apr 19 2013 | 2:28 pm
      Yeah I know! He's just finished his doctorate, and a substantial portion of his thesis relates to contour recognition, based on input from discrete strings' audio output.
      I'm living vicariously of course.
    • Apr 28 2013 | 7:40 pm
      Hi Richard
      any joy from Ricky?
    • May 01 2013 | 9:33 am
      Hi Brendan,
      yes, we've been in touch and he's offered me some pointers. I've got something working now though the whole patch is a bit flakey.Still, it held together yesterday when I played the guitar piece to an audience for the first time.
      Barry Truax was sitting about four feet away from me - so no pressure there then.
      Anyway it went OK- now to make it work properly.
    • May 01 2013 | 10:58 am
      Glad to hear it - at least the performance anxiety was based in illustrious peer pressure and not unreliable patch fear.