Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Incorrect documentation for librosa.core.spectrum.istft() function?

See original GitHub issue

I’m not sure if this is an actual “issue” or just an indication of my lack of understanding, but it seems like the documentation for librosa.core.spectrum.istft() may be incorrect.

The documentation in the code indicates that the function operates by minimizing mean squared error as described in Griffin & Lim. But I don’t think that’s what is going on in this function.

Issue Analytics

State:
Created 5 years ago
Comments:5 (4 by maintainers)

Top GitHub Comments

1reaction

lostanlencommented, Aug 7, 2018

What is indeed puzzling in the documentation is that it cites the Griffin-Lim paper, which is mostly reknowned today as a phase retrieval paper, with the algorithm in Figure 1 (“the Griffin-Lim algorithm”). Yet, librosa.core.spectrum.istft only uses the elements of the paper up to Section 2 ("reconstruction from MSTFT). It might be good to reference Section 2 specifically.

If you tag this issue as 0.6.3 I can write a spiel about that in the docs.

1reaction

r9y9commented, Jun 9, 2018

Reading Griffin & Lim’s paper (PDF link) again, I think it’s correct. See section 2 of their paper (eq. (4) ~ (6)). Any suggestions for wording are welcome though.

MSE minimization is not about phase retrieval, which might be confusing since Grifin & lim algorithm often reffered as a phase reconstruction algorithm.

Top Results From Across the Web

librosa.istft — librosa 0.10.0.dev0 documentation

Converts a complex-valued spectrogram stft_matrix to time-series y by minimizing the mean squared error between stft_matrix and STFT of y as described in...

Source code for librosa.core.spectrum

This function returns a complex-valued matrix D such that - ``np.abs(D[..., f, ... [docs]@cache(level=30) def istft( stft_matrix, *, hop_length=None, ...

Source code for librosa.core.spectrum

This function returns a complex-valued matrix D such that ... [docs]@cache(level=30) def istft( stft_matrix: np.ndarray, ... S = librosa.core.spectrum.

librosa.stft — librosa 0.10.0.dev0 documentation

The STFT represents a signal in the time-frequency domain by computing discrete Fourier transforms (DFT) over short overlapping windows. This function returns a ......

librosa.griffinlim — librosa 0.10.0.dev0 documentation

windowstring, tuple, number, function, or np.ndarray [shape=(n_fft,)]. A window specification as supported by stft or istft. centerboolean.