-1

Hi I am hoping to make an android application that can listen to different whistles of a person (long whistle long whistle short whistle for example) and compare it with a prerecorded whistle of the same type by a different person and determine if they whistled the same way.

I was hoping to use the library musicg for this but I'm unsure whether it would be possible before I even tried? Unfortunately I'm also having trouble getting the demo to work which seems to be an issue in the issue log of the site, but I haven't found any answers yet.

Cheers for any help, apologies for the vague question.

John McKenzie
  • 420
  • 6
  • 16
  • Welcome to StackOverflow. Your question is a bit broad. Better to have more a concrete question. I have no idea if this library is sufficient for your needs, but if you had a problem building/running the sample did you look through all of the issues on the project page to try to make the library work? – Morrison Chang Aug 31 '14 at 18:36
  • Yeah I checked it out already on the issue page but unfortunately I don't see any fixes unless I missed it. the closest I can find is from here [fix](http://stackoverflow.com/questions/14767476/detect-whistle-sound-in-android) but trying to test it now. – John McKenzie Aug 31 '14 at 18:50
  • On StackOverflow we usually expect you to have tried before asking the question. – marko Aug 31 '14 at 20:32
  • A cursory look at the source code for this shows it to be incomplete (for instance, lack of a re-sampling filter in the re-sampler) and naively/inefficiently implemented (FFTs implementing using double precision floating point, and not even a particularly efficient FFT algorithm, and in Java - the performance of this will be terrible on any phone). I'd suggest plundering [Sonic Visualiser](http://www.sonicvisualiser.org) for ideas. It's C++, but you could probably get bits of it to build with the NDK. – marko Aug 31 '14 at 20:49
  • In respect to the tested myself part, apologies but I sometimes get ahead of myself. I pretty much wanted to compare 2 sound files and see if they are similar. But the app which is broken only allows detection of whistles, so I was wondering if it also did a comparison or if there was a better suited lib. For the other part it seems you are right, I ended up finding the comparison bit and as you say its quite inefficient, but I didn't need it to do much either so it works for what I need it for. Thanks heaps for your answers :D – John McKenzie Aug 31 '14 at 21:52

1 Answers1

0

So I ended up just diving into musicg and seeing how it go's. Turns out you can compare the foot print of two sound files quite easily, and it even outputs a float value in terms of how likely to be similar it is. As someone has said above its quite slow, especially on a phone, but works great for what I need it for. Thanks guys for your help.

John McKenzie
  • 420
  • 6
  • 16