python pYIN
Pitch and note tracking in monophonic (a cappella) audio
This is a fork of the python version of pYIN of Matthias Mauch originally ported here
In this fork, following modifications of the original implementation are made:
-
we add efficient numpy data structures for HMM-based note-segmentation (2. step of pYIN)
-
the transition model for note tracking is adapted to take into account positions in the musical measure/bar. This is activated by the flag WITH_BEAT_ANNOS More specifically, we use likelihoods of note onset events (in
code.MonoNoteParameters.barPositionDist_Probs), at different bar position (e.g. from 0 to 9, depending on the bar). In the original version the transition likelihood from silence to a following note attack state is distributed by the pitch difference from current to following note. NOTE: the sum of the transition likelihoods over all possible following notes is a constant1-selfSilenceTransition. In this version, on decoding, this constant is replaced for each time frame by the likelihood of the closest bar position frombarPositionDist_Probs. This means essentially, that the same pitch-difference distribution scheme is kept, but scaled varyingly when close in time to a beginning of the bar (e.g. scaled more at downbeats and less else). -
the pYIN pitch tracking is replaced by predominantmelodymakam. The method is in AlignmentDuration If you want to eliminate the dependence on it, simply call another pitch extraction algorithm or set WITH_MELODIA=0
-
for efficient Viterbi decoding numpy is used in the class https://github.com/georgid/pypYIN/blob/master/pypYIN/MonoNoteHMM.py
https://code.soundsoftware.ac.uk/projects/pyin
Numpy
Scipy
Essentia
https://github.com/craffel/mir_eval
python demo.py
Copyright (C) 2017 Music Technology Group - Universitat Pompeu Fabra
This file is part of pypYIN
pypYIN is free software: you can redistribute it and/or modify it under
the terms of the GNU Affero General Public License as published by the Free
Software Foundation (FSF), either version 3 of the License, or (at your
option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
FOR A PARTICULAR PURPOSE. See the GNU General Public License for more
details.
You should have received a copy of the Affero GNU General Public License
version 3 along with this program. If not, see http://www.gnu.org/licenses/
Georgi Dzhambazov, André Holzapfel, Ajay Srinivasamurthy, Xavier Serra, Metrical-Accent Aware Vocal Onset Detection in Polyphonic Audio, In Proceedings of ISMIR 2017
If you have any problem about this python version code, please contact: georgi.dzhambazov@upf.edu
If you have any problem about this algorithm, I suggest you to contact: Matthias Mauch
m.mauch@qmul.ac.uk who is the original C++ version author of this algorithm or consider the paper
M. Mauch, C. Cannam, R. Bittner, G. Fazekas, J. Salamon, J. Dai, J. Bello and S. Dixon,
“Computer-aided Melody Note Transcription Using the Tony Software: Accuracy and Efficiency”,
in Proceedings of the First International Conference on Technologies for
Music Notation and Representation, 2015.