A collection of papers and resources related to frequency-domain monaural speech enhancement.
*Note: CTSNet, G2Net, and TaylorSENet have been updated.
When using the models provided in this website, please refer to our survey:
Chengshi Zheng#*, Huiyong Zhang, Wenzhe Liu, Xiaoxue Luo#, Andong Li, Xiaodong Li, and Brian C. J. Moore#. Sixty Years of Frequency-Domain Monaural Speech Enhancement: From Traditional to Deep Learning Methods. Trends in Hearing. 2023;27. doi:10.1177/23312165231209913
Please let us know if you find errors or have suggestions to improve the quality of this project by sending an email to: cszheng@mail.ioa.ac.cn; luoxiaoxue@mail.ioa.ac.cn
@article{ZhengTIH2023_Survey,
author = {Chengshi Zheng and Huiyong Zhang and Wenzhe Liu and Xiaoxue Luo and Andong Li and Xiaodong Li and Brian C. J. Moore},
title ={Sixty Years of Frequency-Domain Monaural Speech Enhancement: From Traditional to Deep Learning Methods},
journal = {Trends in Hearing},
volume = {27},
number = {},
pages = {23312165231209913},
year = {2023},
doi = {10.1177/23312165231209913}
}
Paper Download link: https://journals.sagepub.com/doi/full/10.1177/23312165231209913
Introduction
This survey paper first provides a comprehensive overview of traditional and deep-learning methods for monaural speech enhancement in the frequency domain. The fundamental assumptions of each approach are then summarized and analyzed to clarify their limitations and advantages. A comprehensive evaluation of some typical methods was conducted using the WSJ + DNS and Voice Bank + DEMAND datasets to give an intuitive and unified comparison. The benefits of monaural speech enhancement methods using objective metrics relevant for normal-hearing and hearing-impaired listeners were evaluated.
Available models
Results
- Objective test results using the Voice Bank + DEMAND dataset when the input feature was uncompressed. Best scores are highlighted in Bold.
- Objective test results using the Voice Bank + DEMAND dataset when the input feature was compressed. Best scores are highlighted in Bold.
- Values of the HASQI (%)/HASPI (%) for the different methods using the Voice Bank + DEMAND dataset. For all deep-learning methods, both the uncompressed spectrum and the compressed spectrum were used. Bold font indicates the best average score in each group.
Citation guide
[1] Nicolson A and Paliwal KK (2019) Deep learning for minimum mean-square error approaches to speech enhancement. Speech Communication 111: 44–55. DOI: 10.1016/j.specom.2019.06.002.
[2] Sun L, Du J, Dai LR and Lee CH (2017) Multiple-target deep learning for LSTM-RNN based speech enhancement. In: 2017 Hands-free Speech Communications and Microphone Arrays (HSCMA). pp. 136–140. DOI:10.1109/HSCMA.2017. 7895577.
[3] Hao X, Su X, Horaud R and Li X (2021) Fullsubnet: A full-band and sub-band fusion model for real-time single-channel speech enhancement. In: 2021 IEEE International Conference on Acoustics, Speech and Signal Processing. pp. 6633–6637. DOI: 10.1109/ICASSP39728.2021.9414177.
[4] Tan K and Wang D (2018) A convolutional recurrent neural network for real-time speech enhancement. In: Proc. Interspeech 2018. pp. 3229–3233. DOI:doi:10.21437/Interspeech.2018-1405.
[5] Tan K and Wang D (2020) Learning complex spectral mapping with gated convolutional recurrent networks for monaural speech enhancement. IEEE/ACM Transactions on Audio, Speech, and Language Processing 28: 380–390. DOI:10.1109/TASLP. 2019.2955276.
[6] Le X, Chen H, Chen K and Lu J (2021) DPCRN: Dualpath convolution recurrent network for single channel speech enhancement. arXiv preprint arXiv:2107.05429.
[7] Fu Y, Liu Y, Li J, Luo D, Lv S, Jv Y and Xie L (2022) Uformer: A Unet based dilated complex and real dual-path conformer network for simultaneous speech enhancement and dereverberation. In: 2022 IEEE International Conference on Acoustics, Speech and Signal Processing. pp. 7417–7421. DOI: 10.1109/ICASSP43922.2022.9746020.
[8] Hu Y, Liu Y, Lv S, Xing M, Zhang S, Fu Y, Wu J, Zhang B and Xie L (2020) DCCRN: Deep complex convolution recurrent network for phase-aware speech enhancement. arXiv preprint arXiv:2008.00264.
[9] Li A, Liu W, Luo X, Zheng C and Li X (2021b) ICASSP 2021 Deep Noise Suppression Challenge: Decoupling magnitude and phase optimization with a two-stage deep network. In: ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). pp. 6628–6632. DOI: 10.1109/ICASSP39728.2021.9414062.
[10] Li A, Zheng C, Zhang L and Li X (2022b) Glance and gaze: A collaborative learning framework for single-channel speech enhancement. Applied Acoustics 187: 108499. DOI:https: //doi.org/10.1016/j.apacoust.2021.108499
[11] Li A, You S, Yu G, Zheng C and Li X (2022a) Taylor, can you hear me now? a Taylor-unfolding framework for monaural speech enhancement. In: Raedt LD (ed.) Proceedings of the Thirty-First International Joint Conference on Artificial Intelligence, IJCAI-22. International Joint Conferences on Artificial Intelligence Organization, pp. 4193–4200. DOI: 10.24963/ijcai.2022/582. Main Track.