Data Augmentation and Loss Normalization for Deep Noise Suppression

By Sebastian Braun, Ivan J. Tashev. 2020

90 citations recorded by the source corpus.

Speech enhancement using neural networks is recently receiving large attention in research and being integrated in commercial devices and applications. In this work, we investigate data augmentation techniques for supervised deep learning-based speech enhancement. We show that not only augmenting SNR values to a broader range and a continuous distribution helps to regularize training, but also augmenting the spectral and dynamic level diversity. However, to not degrade training by level augmentation, we propose a modification to signal-based loss functions by applying sequence level normalization. We show in experiments that this normalization overcomes the degradation caused by training on sequences with imbalanced signal levels, when using a level-dependent loss function.

Subfields: Computational Mechanics, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Mechanics

In Signal Processing

Citations in the corpus

Icassp 2022 Deep Noise Suppression Challenge (2022, methodology): “We trained NSNet2 [14, 15] on fullband non-personalized training dataset to obtain the baseline.”
Harmonic Gated Compensation Network Plus for ICASSP 2022 DNS Challenge (2022, methodology): “835 framework [26] and Word Accuracy (WAcc) are used to evaluate the speech quality, and HGCN+ outperforms the baseline [2] with less parameters in all metrics, as shown in Table 3.”
ICASSP 2021 Deep Noise Suppression Challenge (2020, methodology): “We also provided the baseline noise suppressor [30] for the participants to benchmark their methods.”