License clarification

[pierre-haessig]

Hi,

I discovered this algorithm thanks to http://dsp.stackexchange.com/questions/129/how-can-i-compute-a-log-spaced-power-spectrum and I am interested in using it.
Since I'm working with Python, this will require a reimplementation but I need to be sure of the license. Is your code available under a BSD license ?

best,
Pierre

[tobin]

Dear Pierre,
Yes, please feel free to use my code under the BSD license. The actual LPSD algorithm, however, has been published in a scientific journal and, as far as I know, there are no restrictions on its re-implementation. So I think you should able to freely reimplement it in Python and do what you want with the resulting code.
Tobin

[fommil]

Was just about to ask the same!

Could you please release under Apache 2.0, which is basically BSD but clearing up any ambiguity around patents held by the author.

[fommil]

@pierre-haessig if you've implemented in python (e.g. as an alternative to numpy.fft.rfft with rescaling), I'd very much like to see that! 😄

[pierre-haessig]

@fommil I have no reimplementation. I only had a crude wrapper around matplotlib.mlab.psd which select NFFT depending on the frequency band.

However, I just took a quick look at the original article http://www.sciencedirect.com/science/article/pii/S026322410500117X and they report problems in preexisting methods (like discontinuity in noise level in Fig 9) which I also faced. So their approach looks much better.

Anyway, I've not been working on spectrums this year, but if I do I should remember to take a furter look at this work!

[pierre-haessig]

Also, coming back to the original issue, I think it can be closed as "fixed": code is available as BSD. Ideally, a license file would be added to the repo to make things clear.

Thanks @tobin for sharing your work!

[rudolfbyker]

I've started work on a Python translation of this, but since I don't know Matlab, it's taking a while. Work in Progress here: https://github.com/rudolfbyker/lpsd

[rudolfbyker]

I think it's done. Are any of you interested in helping to test whether my Python implementation is equivalent to the Matlab implementation? See: rudolfbyker#1