Learning to Denoise Historical Music
Authors:
- Yunpeng LiYLYunpeng Li
- Beat GfellerBGBeat Gfeller
- Marco TagliasacchiMTMarco Tagliasacchi
- Dominik RoblekDRDominik Roblek
Preprint
Abstract
We propose an audio-to-audio neural network model that learns to denoise old music recordings. Our model internally converts its input into a time-frequency representation by means of a short-time Fourier transform (STFT), and processes the resulting complex spectrogram using a convolutional neural network. The network is trained with both reconstruction and adversarial objectives on a synthetic noisy music dataset, which is created by mixing clean music with real noise samples extracted from quiet segments of old recordings. We evaluate our method quantitatively on held-out test examples of the synthetic dataset, and qualitatively by human rating on samples of actual historical recordings. Our results show that the proposed method is effective in removing noise, while preserving the quality and details of the original music.Keywords
All Related Versions
This publication has 14 references indexed in Scilit:
- SciPy 1.0: fundamental algorithms for scientific computing in PythonNature Methods, 2020
- MedGAN: Medical image translation using GANsComputerized Medical Imaging and Graphics, 2020
- Exploiting time-frequency patterns with LSTM-RNNs for low-bitrate audio restorationNeural Computing and Applications, 2019
- Exploring Speech Enhancement with Generative Adversarial Networks for Robust Speech Recognition2018
- Inpainting of Long Audio Segments With Similarity GraphsIEEE/ACM Transactions on Audio, Speech, and Language Processing, 2018
- U-Net: Convolutional Networks for Biomedical Image Segmentation2015
- Sparse Linear Regression With Structured Priors and Application to Denoising of Musical AudioIEEE Transactions on Audio, Speech, and Language Processing, 2008
- Soft Mask Methods for Single-Channel Speaker SeparationIEEE Transactions on Audio, Speech, and Language Processing, 2007
- Speech enhancement based on perceptually motivated bayesian estimators of the magnitude spectrumIEEE Transactions on Speech and Audio Processing, 2005
- Speech enhancement using a minimum mean-square error log-spectral amplitude estimatorIEEE Transactions on Acoustics, Speech, and Signal Processing, 1985