This is the onset detection method described in these publications:
"Note onset detection using one semitone filter-bank for MIREX 2009",
A. Pertusa and J. M. Iñesta,
in MIREX 2009 audio onset detection contest,
http://www.music-ir.org/mirex/2009/index.php/Audio_Onset_Detection_Results
"Recognition of note onsets in digital music using semitone bands",
A. Pertusa, A. Klapuri and J. M. Iñesta,
Lecture notes in Computer Science, vol. 3773, pp 869-879, 2005.
The input file must be WAV, MONO, and with fs= 22050 Hz or 44100 Hz. The default output is a list of onset times in seconds printed in the standard output.
To compile the algorithm in Linux or MacOS, type from the command line:
make
This software requires the libsndfile and fftw3 libraries. Both are standard in Linux distributions and they can be obtained from the corresponding package installer. In MacOS, they can be installed using darwinports. Alternatively, these libraries can be downloaded from:
http://www.mega-nerd.com/libsndfile/ http://www.fftw.org/
The basic syntax once compiled is:
./onset input.wav > onsets.txt
Some parameters can also be controlled from command line:
./onset input.wav [-c context | -t theta | -m mu | -csound | -o odffile ] > onsets.txt
where context is the number of aditional frames, theta is the onset detection threshold, and mu is the silence threshold. The default values are context=0, theta=0.18 and mu=70.
The -csound option generates two files (onset.orc and onset.sco). They contain the instructions to generate (using CSound) synthesized wavefiles with added "ticks" in the times when onsets were detected. CSound comes as a standard package in Linux and MacOS. It can also be downloaded from www.csounds.com
The -o <filename> option prints the onset detection function into the specified file.