mlsanityData to 19 Oct 2026

A study of the robustness of raw waveform based speaker embeddings under mismatched conditions

arXiv 2110.04265 ↗First posted 8 Oct 2021eess.AS cs.SDcited by 0

On the map: Speech and audio language models → Speaker recognition and diarization

Abstract (arXiv)

In this paper, we conduct a cross-dataset study on parametric and non-parametric raw-waveform based speaker embeddings through speaker verification experiments. In general, we observe a more significant performance degradation of these raw-waveform systems compared to spectral based systems. We then propose two strategies to improve the performance of raw-waveform based systems on cross-dataset tests. The first strategy is to change the real-valued filters into analytic filters to ensure shift-invariance. The second strategy is to apply variational dropout to non-parametric filters to prevent them from overfitting irrelevant nuance features.

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 16 of its 26 references that are in the map. First posted on arXiv 8 Oct 2021.

Builds on

16 of 26 references.

References · 16

  1. 2018VoxCeleb2: Deep Speaker Recognitioncited by 1,166
  2. 2017VoxCeleb: a large-scale speaker identification datasetcited by 985
  3. 2019PANNs: Large-Scale Pretrained Audio Neural Networks for Audio Pattern Recognitioncited by 603
  4. 2018Conv-TasNet: Surpassing Ideal Time-Frequency Magnitude Masking for Speech Separationcited by 596
  5. 2019Making Convolutional Networks Shift-Invariant Againcited by 395
  6. 2017Variational Dropout Sparsifies Deep Neural Networkscited by 266
  7. 2018Speaker Recognition from Raw Waveform with SincNetcited by 205
  8. 2021LEAF: A Learnable Frontend for Audio Classificationcited by 79
  9. 2018Voices Obscured in Complex Environmental Settings (VOICES) corpuscited by 47
  10. 2017Learning Filterbanks from Raw Speech for Phone Recognitioncited by 27
  11. 2020Improved RawNet with Feature Map Scaling for Text-independent Speaker Verification using Raw Waveformscited by 24
  12. 2019Filterbank design for end-to-end speech separationcited by 22
4 more
  1. 2017Sample-level CNN Architectures for Music Auto-tagging Using Raw Waveformscited by 14
  2. 2020Interpretable Representation Learning for Speech and Audio Signals Based on Relevance Weightingcited by 7
  3. 2020Y-Vector: Multiscale Waveform Encoder for Speaker Embeddingcited by 6
  4. 2021Learning Sparse Analytic Filters for Piano Transcriptioncited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2022Scoring of Large-Margin Embeddings for Speaker Verification: Cosine or PLDA?
  2. 2025Interpreting the Dimensions of Speaker Embedding Space
  3. 2022DT-SV: A Transformer-based Time-domain Approach for Speaker Verification
  4. 2022Analyzing speaker verification embedding extractors and back-ends under language and channel mismatch
  5. 2024Challenging margin-based speaker embedding extractors by using the variational information bottleneck
  6. 2021A Speaker Verification Backend with Robust Performance across Conditions
  7. 2026Towards Robust Uncertainty-Aware Speaker Modeling
  8. 2022Pushing the limits of raw waveform speaker recognition