A repository and webpage accompanying the paper Janssen 2.0: Audio Inpainting in the Time-frequency Domain.
The paper focuses on inpainting missing parts of an audio signal spectrogram, i.e., estimating the lacking time-frequency coefficients. The autoregression-based Janssen algorithm, a state-of-the-art for the time-domain audio inpainting, is adapted for the time-frequency setting. This novel method, termed Janssen-TF, is compared with the deep-prior neural network approach using both objective metrics and a subjective listening test, proving Janssen-TF to be superior in all the considered measures.
The paper compares a recent method abbreviated DPAI with the newly proposed Janssen-TF approach.
- DPAI codes are not a part of this repository but are available here.
- Matlab codes of our method are available in the
Janssen-TFfolder. - For reproducibility reasons, the codes are set to read the input (uncorrupted) audio files from the
audio-originalsfolder. - The spectrogram masks used in our experiments are read from the
masksfolder. - Regarding the mid-scale experiment using the IRMAS dataset, the folder
audio-irmasincludes a list of the files used in our experiment and a Matlab script which crops the files to a length of 5 seconds and subsamples them to 16 kHz. The original files can be downloaded here.
Note that there are several autoregression-based methods implemented in the Janssen-TF folder;
to exactly reproduce results from the paper, switch to ADMM, primal.
This provides the time-domain signal from line 7 of the algorithm in the paper, after the convergence is reached.
The Matlab codes for Janssen-TF use the LTFAT and the Signal Processing Toolbox. To compute the perceptually-motivated evaluation, we have used the PEMO-Q package (version 1.4.1).