# Gradient weighting for speaker verification in extremely low Signal-to-Noise Ratio

arXiv: https://arxiv.org/abs/2401.02626 · HTML: https://mlsanity.com/paper/2401.02626/

## Key facts

- No citing papers in the map yet (data through 19 Oct 2026).
- Placed in Speaker recognition and diarization, one of 11 directions in Speech and audio language models.
- Builds on 22 of its 29 references that are in the map.
- First posted on arXiv 5 Jan 2024.

## Placed in

- Speech and audio language models / [Speaker recognition and diarization](https://mlsanity.com/speech/speaker-recognition-and-diarization/)

## Builds on (most cited first, 22 of 22)

- [Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization](https://mlsanity.com/paper/1610.02391/) (2016)
- [Axiomatic Attribution for Deep Networks](https://mlsanity.com/paper/1703.01365/) (2017)
- [VoxCeleb2: Deep Speaker Recognition](https://mlsanity.com/paper/1806.05622/) (2018)
- [MUSAN: A Music, Speech, and Noise Corpus](https://mlsanity.com/paper/1510.08484/) (2015)
- [LLM-Pruner: On the Structural Pruning of Large Language Models](https://mlsanity.com/paper/2305.11627/) (2023)
- [DeepCache: Accelerating Diffusion Models for Free](https://mlsanity.com/paper/2312.00858/) (2023)
- [Real Time Image Saliency for Black Box Classifiers](https://mlsanity.com/paper/1705.07857/) (2017)
- [The INTERSPEECH 2020 Deep Noise Suppression Challenge: Datasets, Subjective Testing Framework, and Challenge Results](https://mlsanity.com/paper/2005.13981/) (2020)
- [Wespeaker: A Research and Production oriented Speaker Embedding Learning Toolkit](https://mlsanity.com/paper/2210.17016/) (2022)
- [Deep Model Reassembly](https://mlsanity.com/paper/2210.17409/) (2022)
- [Proper Network Interpretability Helps Adversarial Robustness in Classification](https://mlsanity.com/paper/2006.14748/) (2020)
- [MFA: TDNN with Multi-scale Frequency-channel Attention for Text-independent Speaker Verification with Short Utterances](https://mlsanity.com/paper/2202.01624/) (2022)
- [VoiceID Loss: Speech Enhancement for Speaker Verification](https://mlsanity.com/paper/1904.03601/) (2019)
- [Disentangling Voice and Content with Self-Supervision for Speaker Recognition](https://mlsanity.com/paper/2310.01128/) (2023)
- [Feature Enhancement with Deep Feature Losses for Speaker Verification](https://mlsanity.com/paper/1910.11905/) (2019)
- [UNetGAN: A Robust Speech Enhancement Approach in Time Domain for Extremely Low Signal-to-noise Ratio Condition](https://mlsanity.com/paper/2010.15521/) (2020)
- [Extended U-Net for Speaker Verification in Noisy Environments](https://mlsanity.com/paper/2206.13044/) (2022)
- [Towards Robust Speaker Verification with Target Speaker Enhancement](https://mlsanity.com/paper/2103.08781/) (2021)
- [Analysis of Deep Feature Loss based Enhancement for Speaker Verification](https://mlsanity.com/paper/2002.00139/) (2020)
- [Reliable Visualization for Deep Speaker Recognition](https://mlsanity.com/paper/2204.03852/) (2022)
- [PL-EESR: Perceptual Loss Based END-TO-END Robust Speaker Representation Extraction](https://mlsanity.com/paper/2110.00940/) (2021)
- [Visualizing data augmentation in deep speaker recognition](https://mlsanity.com/paper/2305.16070/) (2023)

## Abstract (arXiv)

Speaker verification is hampered by background noise, particularly at extremely low Signal-to-Noise Ratio (SNR) under 0 dB. It is difficult to suppress noise without introducing unwanted artifacts, which adversely affects speaker verification. We proposed the mechanism called Gradient Weighting (Grad-W), which dynamically identifies and reduces artifact noise during prediction. The mechanism is based on the property that the gradient indicates which parts of the input the model is paying attention to. Specifically, when the speaker network focuses on a region in the denoised utterance but not on the clean counterpart, we consider it artifact noise and assign higher weights for this region during optimization of enhancement. We validate it by training an enhancement model and testing the enhanced utterance on speaker verification. The experimental results show that our approach effectively reduces artifact noise, improving speaker verification across various SNR levels.