Clean and degraded signals

clean	clipped	gaps	random loss
	55% samples	40 ms	60% samples

Reconstructions

method	clipped	gaps	random loss
reweighted [1]
SPAIN [2]
Janssen [3]
NMF [4]

Nonnegative matrix factorization

original signal

NMF components
1		2		3		4
5		6		7		8

References

1. O. Mokrý and P. Rajmic, “Audio Inpainting: Revisited and Reweighted,” in IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 28, 2020. 10.1109/TASLP.2020.3030486
2. O. Mokrý, P. Záviška, P. Rajmic and V. Veselý, “Introducing SPAIN (SParse Audio INpainter),” 2019 27th European Signal Processing Conference (EUSIPCO), A Coruna, Spain, 2019. 10.23919/EUSIPCO.2019.8902560 [GitHub]
3. A. J. E. M. Janssen, R. N. J. Veldhuis, and L. B. Vries, “Adaptive interpolation of discrete-time signals that can be modeled as autoregressive processes,” IEEE Transactions on Acoustics, Speech and Signal Processing, vol. 34, no. 2, 1986. 10.1109/TASSP.1986.1164824
4. O. Mokrý, P. Magron, T. Oberlin and C. Févotte, “Algorithms for audio inpainting based on probabilistic nonnegative matrix factorization,” Signal Processing, vol 206, 2023. 10.1016/j.sigpro.2022.108905 [GitHub]