Advanced search
1 file | 2.35 MB Add to list

A novel channel estimate for noise robust speech recognition

Geoffroy Vanderreydt (UGent) and Kris Demuynck (UGent)
Author
Organization
Abstract
We propose a novel technique to estimate the channel characteristics for robust speech recognition. The method focuses on reliable time-frequency speech patches which are highly independent of the noise condition. Combined with a root -based approximation of the logarithm in the MFCC computation, this reduces the variance caused by the noise on the spectral features, and therefore also the constrain on the acoustic model in a multi -style training setup. We show that compared to the standard mean normalization, the proposed method estimates the channel equally well under clean conditions and better under noisy conditions. When integrated in the feature extraction pipeline, we show improvements in speech recognition accuracy on noisy speech and a status quo on clean speech. Our experiments reveal that this method helps the most for generative models that need to model the complex noise variability, and less so for discriminative models, which can learn to ignore noise instead of accurately modeling it. Our approach outperforms the state of the art on the noisy Aurora4 task.
Keywords
Channel estimate, Noise robustness, Speech recognition, Feature engineering, AUTOMATIC SPEECH, NEURAL-NETWORKS, FEATURES, ADAPTATION

Downloads

  • (...).pdf
    • full text (Published version)
    • |
    • UGent only
    • |
    • PDF
    • |
    • 2.35 MB

Citation

Please use this url to cite or link to this publication:

MLA
Vanderreydt, Geoffroy, and Kris Demuynck. “A Novel Channel Estimate for Noise Robust Speech Recognition.” COMPUTER SPEECH AND LANGUAGE, vol. 86, 2024, doi:10.1016/j.csl.2023.101598.
APA
Vanderreydt, G., & Demuynck, K. (2024). A novel channel estimate for noise robust speech recognition. COMPUTER SPEECH AND LANGUAGE, 86. https://doi.org/10.1016/j.csl.2023.101598
Chicago author-date
Vanderreydt, Geoffroy, and Kris Demuynck. 2024. “A Novel Channel Estimate for Noise Robust Speech Recognition.” COMPUTER SPEECH AND LANGUAGE 86. https://doi.org/10.1016/j.csl.2023.101598.
Chicago author-date (all authors)
Vanderreydt, Geoffroy, and Kris Demuynck. 2024. “A Novel Channel Estimate for Noise Robust Speech Recognition.” COMPUTER SPEECH AND LANGUAGE 86. doi:10.1016/j.csl.2023.101598.
Vancouver
1.
Vanderreydt G, Demuynck K. A novel channel estimate for noise robust speech recognition. COMPUTER SPEECH AND LANGUAGE. 2024;86.
IEEE
[1]
G. Vanderreydt and K. Demuynck, “A novel channel estimate for noise robust speech recognition,” COMPUTER SPEECH AND LANGUAGE, vol. 86, 2024.
@article{01HS0MRXFGYW95TTKEWPS027HP,
  abstract     = {{We propose a novel technique to estimate the channel characteristics for robust speech recognition. The method focuses on reliable time-frequency speech patches which are highly independent of the noise condition. Combined with a root -based approximation of the logarithm in the MFCC computation, this reduces the variance caused by the noise on the spectral features, and therefore also the constrain on the acoustic model in a multi -style training setup. We show that compared to the standard mean normalization, the proposed method estimates the channel equally well under clean conditions and better under noisy conditions. When integrated in the feature extraction pipeline, we show improvements in speech recognition accuracy on noisy speech and a status quo on clean speech. Our experiments reveal that this method helps the most for generative models that need to model the complex noise variability, and less so for discriminative models, which can learn to ignore noise instead of accurately modeling it. Our approach outperforms the state of the art on the noisy Aurora4 task.}},
  articleno    = {{101598}},
  author       = {{Vanderreydt, Geoffroy and Demuynck, Kris}},
  issn         = {{0885-2308}},
  journal      = {{COMPUTER SPEECH AND LANGUAGE}},
  keywords     = {{Channel estimate,Noise robustness,Speech recognition,Feature engineering,AUTOMATIC SPEECH,NEURAL-NETWORKS,FEATURES,ADAPTATION}},
  language     = {{eng}},
  pages        = {{15}},
  title        = {{A novel channel estimate for noise robust speech recognition}},
  url          = {{http://doi.org/10.1016/j.csl.2023.101598}},
  volume       = {{86}},
  year         = {{2024}},
}

Altmetric
View in Altmetric
Web of Science
Times cited: