This is an website accompanying three articles about Audio Declipping (see the abstracts below). On this website you can find supplementary material such as a detailed description of the audio dataset used, additional plots with results for each individual audio excerpt, a link to the repository with MATLAB source codes, and last but not least, you can listen to the restored audio excerpts.

Abstract: Dynamic range limitations in signal processing often lead to clipping, or saturation, in signals. Audio declipping is the task of estimating the original audio signal given its clipped measurements and has attracted a lot of interest in recent years. Audio declipping algorithms often make assumptions about the underlying signal, such as sparsity or low-rankness, as well as the measurement system. In this paper, we provide an extensive review of audio declipping algorithms proposed in the literature. For each algorithm, we present the assumptions being made about the audio signal, the modeling domain, as well as the optimization algorithm. Furthermore, we provide an extensive numerical evaluation of popular declipping algorithms, on real audio data. We evaluate each algorithm in terms of the Signal-to-Distortion Ratio, as well as using perceptual metrics of sound quality. The article is accompanied with the repository containing the evaluated methods.

Full-text: IEEE Xplore, arXiv postprint.
Citations: Plain Text, BibTeX.

Abstract: Some audio declipping methods produce waveforms that do not fully respect the physical process of clipping, which is why we refer to them as inconsistent. This letter reports what effect on perception it has if the solution by inconsistent methods is forced consistent by postprocessing. We first propose a simple sample replacement method, then we identify its main weaknesses and propose an improved variant. The experiments show that the vast majority of inconsistent declipping methods significantly benefit from the proposed approach in terms of objective perceptual metrics. In particular, we show that the SS PEW method based on social sparsity combined with the proposed method performs comparable to top methods from the consistent class, but at a computational cost of one order of magnitude lower.

Full-text: ScienceDirect, arXiv preprint.
Citations: Plain Text, BibTeX.

Abstract: We develop the analysis (cosparse) variant of the popular audio declipping algorithm of Siedenburg et al. (2014). Furthermore, we extend both the old and the new variants by the possibility of weighting the time-frequency coefficients. We examine the audio reconstruction performance of several combinations of weights and shrinkage operators. The weights are shown to improve the reconstruction quality in some cases; however, the best scores achieved by the non-weighted methods are not surpassed with the help of weights. Yet, the analysis Empirical Wiener (EW) shrinkage was able to reach the quality of a computationally more expensive competitor, the Persistent Empirical Wiener (PEW). Moreover, the proposed analysis variant incorporating PEW slightly outperforms the synthesis counterpart in terms of an auditorily motivated metric.

Full-text: IEEE Xplore, arXiv postprint.
Citations: Plain Text, BibTeX.

Following the idea of reproducible research, we make all the implementations freely available at the GitHub repository.
Please note that LTFAT toolbox (version>=2.4.0, available here) must be installed and loaded in order to run the scripts and reproduce the results.

The following table contains abbreviations and full names of the algorithms used in the evaluation.
Algorithms inconsistent in the reliable part of the clipped signal are marked with an asterisk *.
Note that the analysis variant of Social Sparsity declipper (ASS) was introduced later in the above-mentioned conference paper and thus it is not part of the Declipping Survey nor the Crossfading article.