This python code intend to produce a sine modelling (partial tracking) on spectrograms generated from sounds.
Originally it was designed in an ecoacoustic study, with the corresponding scientific paper below. It can also be used also to produce sine modelling on other types of sounds, like speech.
A new branch provide a python 3 version.
Guyot, P., Eldridge, A., Eyre-Walker, Y. C., Johnston, A., Pellegrini, T., & Peck, M. Sinusoidal modelling for ecoacoustics. In Annual conference Interspeech (INTERSPEECH 2016, pp-2602-2606, link).
It uses a sinusoidal modelling algorithm to compute some audio indices in the context of Ecoacoustic.
It can output:
- a csv files containing the indices,
- a pickle file containing the values of all the tracks,
- a plot of the sinusoidal modelling.
This code shows a version of a tool in development. Further improvements could be added in the future, in terms of computational cost, precision of the results, and documentation.
Remark: some synthesized audio files have been computed (in the "Synth_audio_from_pkl" folder) from the outputted plk files, using an adaptation of the sms-tools for the synthesis (that is not currently available in this repository).
This code is based on python (tested with python 2.7.8).
Some python package are required:
- Numpy
- Scipy
- Matlplotlib (for graphic)
- csv, pickle (to write output files)
- os
Update the values of the variable in the 'main' part of the code. Then run:
python main.py
GPL V 3 - see LICENSE.txt for more info
Patrice Guyot (code adapted from: M. Le Coz, J. Pinquier, and R. André-Obrecht, “Superposed speech localisation using frequency tracking,” in INTERSPEECH 2013 – 13th Annual Conference of the International Speech Communication Association, August 25-29, Lyon, France, Proceedings, 2013.)
Questions, comments and remarks (by emails) would be appreciated.
Credits: Patrice Guyot, Alice Eldridge, Mika Peck