Janssen 2.0: Audio Inpainting in the Time-frequency Domain

TL;DR

This paper introduces Janssen-TF, a novel time-frequency domain audio inpainting method based on the Janssen algorithm, demonstrating superior performance over neural network approaches through objective and subjective evaluations.

Contribution

It adapts the Janssen algorithm for time-frequency domain audio inpainting, providing a new state-of-the-art method that outperforms neural network-based approaches.

Findings

Janssen-TF outperforms neural network methods in objective metrics.

Janssen-TF achieves better subjective listening test results.

The method effectively reconstructs missing parts of audio spectrograms.

Abstract

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.