Hi all, to analyze recordings of environmental noise I'm searching a software (either Windows or Mac, or a R/Matlab/Python package) for calculating 1/3 octave spectral percentiles. Any hint ?
We published an open access paper last year that includes software in MATLAB and R that can do this - default percentiles are 1,5,50,95,99, but I can show you where to modify the code if you need others.
well, there is matlab, octav etc. I think we should go a little bit beyond this special case/question: where do bioacousticians find a sustainable software ecosystem to develop, re-use and find useful packages/subprograms? I wrote code in fortran, visual basic and scripts for various software packages... all gone...:)
Matlab, unfortunately it is expensive. A good package is X-BAT, however I'm not sure it is currently maintained and further developed.
Octave, appears to be compatible with Matlab, but much less powerful for the generation of graphs (maybe I'm wrong...).
R, not very user friendly, with many inconsistencies in the command syntax, many bugs, very slow, much less graphically powerful than Matlab. For R there are two noteworthy packages devoted to bioacousticians: SEEWAVE and SOUNDECOLOGY.
Python, promising, but good packages for bioacoustibcians not yet available.
So, the landscape appears quite poor.
This is my limited "view" and I'd be very happy to be contradicted !
I work on a daily basis swith python and believe me numphy and sciphy are not the best for signal analysis.
The other chance could be making your own Java libraries, but that will require time and money. Better buy a Matlab license and after that pass it to C.
X bat do not have support, and actually its full of bugs.
We're just about to publish a paper on ship noise in which we present quantiles (5, 25, 50, 75, 95%) of source spectrum levels. I doubt the R/ggplot scripts will be as user-friendly as Nathan's resources (above), but you're free to adapt them from this archive --
One of my goals for 2016 is to try to get acoustic analysis workflow optimized in Jupyter (w/Python), so I'm sorry to hear from Mario (above) that signal analysis isn't very far along in Numpy and Scipy...
Audacity may export dB values from an power spectrum. With that in hand you may use excel to sum intensity value within each 1/3 octave band. The same may be done with STx...
I'm receiving conflicting inputs about Python. Many say that is easier and faster than R. I used both and with both I had problems, mostly related with managing libraries and dependancies coming from different authors and with different update levels.
Recently I heard of many researchers willing to move to Python; I also found this interesting paper:
pyAudioAnalysis: An Open-Source Python Library for Audio Signal Analysis
PLOS ONE 10(12):E0144610
To conclude, I'm willing to understand on which programming system I should invest my time and resources !
Not the best does not means that is impossible. What is available for signal analysis is scipy library for SA http://docs.scipy.org/doc/scipy/reference/signal.html.
Probably agree with me that the problem with python is the workload. A good option could be Cython, but please keep us posted would love to see it.