Algorithms (libnmfd.dsp.algorithms)

libnmfd.dsp.algorithms.griffin_lim(X: ndarray, num_iter: int = 50, block_size: int = 2048, hop_size: int = 512, win: ndarray | None = None, append_frames: bool = True, analytic_sig: bool = False, **kwargs) → Tuple[ndarray, ndarray, ndarray][source]

Performs one iteration of the phase reconstruction algorithm as described in [2].

References

[1] Daniel W. Griffin and Jae S. Lim Signal estimation from modified short-time fourier transform IEEE Transactions on Acoustics, Speech and Signal Processing, vol. 32, no. 2, pp. 236-243, Apr 1984.

The operation performs an iSTFT (LSEE-MSTFT) followed by STFT on the resynthesized signal.

Parameters:

  • X (np.ndarray) – The STFT spectrogram to iterate upon

  • num_iter (int) – Number of iterations

  • block_size (int) – The block size to use during analysis

  • hop_size (int) – The used hop size (denoted as S in [1])

  • win (np.ndarray) – Window function

  • append_frames (bool) – If this is enabled, safety spaces have to be removed after the iSTFT

  • analytic_sig (bool) – If this is set to True, we want the analytic signal

Returns:

  • Xout (np.ndarray) – The spectrogram after iSTFT->STFT processing

  • Pout (np.ndarray) – The phase spectrogram after iSTFT->STFT processing

  • res (np.ndarray) – Reconstructed time-domain signal obtained via iSTFT

libnmfd.dsp.algorithms.hpss_kam_fitzgerald(X: ndarray, num_iter: int = 1, kern_dim: int = 17, use_median: bool = False, alpha_param: float = 1.0) → Tuple[List[ndarray], ndarray, int][source]

This re-implements the KAM-based HPSS-algorithm described in [1]. This is a generalization of the median-filter based algorithm first presented in [2]. Our own variant of this algorithm [3] is also supported.

References

[1] Derry FitzGerald, Antoine Liutkus, Zafar Rafii, Bryan Pardo, and Laurent Daudet Harmonic/Percussive Separation using Kernel Additive Modelling Irish Signals and Systems Conference (IET), Limerick, Ireland, 2014, pp. 35�40.

[2] Derry FitzGerald Harmonic/Percussive Separation using Median Filtering In Proceedings of the International Conference on Digital Audio Effects (DAFx), Graz, Austria, 2010, pp. 246-253.

[3] Christian Dittmar, Jonathan Driedger, Meinard Müller, and Jouni Paulus An Experimental Approach to Generalized Wiener Filtering in Music Source Separation In Proceedings of the European Signal Processing Conference (EUSIPCO): 1743–1747, 2016.

Parameters:

  • X (np.ndarray) – Input mixture magnitude spectrogram

  • num_iter (int) – The number of iterations

  • kern_dim (int) – The kernel dimensions

  • use_median (bool) – If True, reverts to FitzGerald’s old method

  • alpha_param (float) – The alpha-Wiener filter exponent

Returns:

  • kam_X (list) – List containing the percussive and harmonic estimate

  • kern (np.ndarray) – The kernels used for enhancing percussive and harmonic part

  • kern_ord (int) – The order of the kernels