A Perceptually-Motivated Approach for Low-Complexity, Real-Time Enhancement of Fullband Speech

Jean-Marc Valin, Umut Isik, Neerad Phansalkar, Ritwik Giri, Karim Helwani, Arvindh Krishnaswamy


Over the past few years, speech enhancement methods based on deep learning have greatly surpassed traditional methods based on spectral subtraction and spectral estimation. Many of these new techniques operate directly in the the short-time Fourier transform (STFT) domain, resulting in a high computational complexity. In this work, we propose PercepNet, an efficient approach that relies on human perception of speech by focusing on the spectral envelope and on the periodicity of the speech. We demonstrate high-quality, real-time enhancement of fullband (48 kHz) speech with less than 5% of a CPU core.


 DOI: 10.21437/Interspeech.2020-2730

Cite as: Valin, J., Isik, U., Phansalkar, N., Giri, R., Helwani, K., Krishnaswamy, A. (2020) A Perceptually-Motivated Approach for Low-Complexity, Real-Time Enhancement of Fullband Speech. Proc. Interspeech 2020, 2482-2486, DOI: 10.21437/Interspeech.2020-2730.


@inproceedings{Valin2020,
  author={Jean-Marc Valin and Umut Isik and Neerad Phansalkar and Ritwik Giri and Karim Helwani and Arvindh Krishnaswamy},
  title={{A Perceptually-Motivated Approach for Low-Complexity, Real-Time Enhancement of Fullband Speech}},
  year=2020,
  booktitle={Proc. Interspeech 2020},
  pages={2482--2486},
  doi={10.21437/Interspeech.2020-2730},
  url={http://dx.doi.org/10.21437/Interspeech.2020-2730}
}