Skip to content
DraconiticResearch
Search Sources API
203,234 studies31 sourcesUpdated 1 min ago
Peer-reviewedCC BY 4.0PLOS One

Speech enhancement method integrating squeeze and excitation channel attention mechanism and improved CGAN

  • Yecai Guo1
  • Meiyu Liang2
  • Tianyue Jiang
  1. 1School of Electrical and Electronic Engineering, Anhui Institute of Information Technology, Wuhu, Anhui, China
  2. 2College of Electronic and Information Engineering, Nanjing University of Information Science and Technology, Nanjing, Jiangsu, China
Open at sourcePDFDOICite

Abstract

To avoid the additional computational cost and algorithmic latency introduced by short-time Fourier transformation and waveform reconstruction, we propose CGAN-SECA, which incorporates a squeeze-and-excitation channel attention mechanism to enhance noisy speech directly in the time domain. The generator employs a symmetric encoder-decoder backbone with skip connections to reconstruct waveform samples and reduce information loss during down-sampling and up-sampling. SECA adaptively recalibrates channel-wise feature responses, enabling the network to emphasize speech-relevant representations under complex noise. In the discriminator, Virtual Batch Normalization (VBN) and Dropout are introduced to regularize adversarial training. Least-squares adversarial objectives replace cross-entropy-based objectives, and an L1 reconstruction term encourages sample-level fidelity. The method is evaluated on VoiceBank-DEMAND, which contains environmental and babble noise, and on the THCHS30-DNS Challenge dataset under multiple Signal-to-Noise Ratio (SNR) conditions. On VoiceBank-DEMAND, CGAN-SECA achieved PESQ and STOI scores of 3.018 and 0.954, respectively, ranking second only to CMGAN and outperforming the remaining comparison methods. CGAN-SECA also maintained fast inference, with a real-time factor (RTF) of 0.00423. On the THCHS30-DNS Challenge dataset, CGAN-SECA achieved the highest average PESQ and STOI scores among all evaluated methods, reaching 2.274 and 0.838, respectively, across the −5, 0, and 5 dB conditions. In particular, it produced the best results for both metrics at 0 and 5 dB, demonstrating effective speech enhancement and good generalization across different noise levels and datasets.

Related

  1. A multi-attention mechanism-improved ResNet18 approach for lithology identification and classification
    Yan Zhang, Fan Zhang, Jiaming Li +2
    PLOS OnePeer-reviewedCC BY 4.0Full textNeuroscienceGeology & geochemistryPLOSPMC
  2. Attention-LSTM Network for Improved Hysteresis Compensation of a Single-Port Surgical Robot
    Xiaohang Zhang, Di Wu, Matthias Tummers +1
    IEEE Transactions on Medical Robotics and BionicsPeer-reviewedLicence Ouverte 2.0Full textSystems & synthetic biologyObstetrics, gynaecology & paediatricsHAL
  3. LeafXNet: A Lightweight Explainable Model With Integrated Attention Mechanism for Fruit Leaf Disease Detection
    Shah Murtaza Rashid Al Masud, Farhana Sultana Eshita, Abu Bakar Siddique Mahi +4
    Plant directPeer-reviewedCC BYFull textPlant biologyEurope PMC
  4. Sentiment classification optimization based on improved Senti-BERT and BiLSTM-Attention
    Jie Zhang, Jianbang Liu, Zhaosheng Xu
    PLOS OnePeer-reviewedCC BY 4.0Full textNeurosciencePLOS
  5. Advancing herbicide safeners: Integrating structure-guided design, pathway mechanism elucidation, and smart delivery
    Xile Deng, Yudan Wu, Lianyang Bai
    Advanced AgrochemPeer-reviewedCC BY7 citedPlant biologyOpenAlex

Cite

BibTeXRIS
BibTeX
@article{guo2026speech,
  title = {{Speech enhancement method integrating squeeze and excitation channel attention mechanism and improved CGAN}},
  author = {Yecai Guo and Meiyu Liang and Tianyue Jiang},
  journal = {PLOS One},
  year = {2026},
  month = oct,
  doi = {10.1371/journal.pone.0359641},
  url = {https://doi.org/10.1371/journal.pone.0359641},
  note = {CC BY 4.0},
}
RIS
TY  - JOUR
TI  - Speech enhancement method integrating squeeze and excitation channel attention mechanism and improved CGAN
AU  - Yecai Guo
AU  - Meiyu Liang
AU  - Tianyue Jiang
PY  - 2026
DA  - 2026/10/01
JO  - PLOS One
DO  - 10.1371/journal.pone.0359641
UR  - https://doi.org/10.1371/journal.pone.0359641
AB  - To avoid the additional computational cost and algorithmic latency introduced by short-time Fourier transformation and waveform reconstruction, we propose CGAN-SECA, which incorporates a squeeze-and-excitation channel attention mechanism to enhance noisy speech directly in the time domain. The generator employs a symmetric encoder-decoder backbone with skip connections to reconstruct waveform samples and reduce information loss during down-sampling and up-sampling. SECA adaptively recalibrates channel-wise feature responses, enabling the network to emphasize speech-relevant representations under complex noise. In the discriminator, Virtual Batch Normalization (VBN) and Dropout are introduced to regularize adversarial training. Least-squares adversarial objectives replace cross-entropy-based objectives, and an L1 reconstruction term encourages sample-level fidelity. The method is evaluated on VoiceBank-DEMAND, which contains environmental and babble noise, and on the THCHS30-DNS Challenge dataset under multiple Signal-to-Noise Ratio (SNR) conditions. On VoiceBank-DEMAND, CGAN-SECA achieved PESQ and STOI scores of 3.018 and 0.954, respectively, ranking second only to CMGAN and outperforming the remaining comparison methods. CGAN-SECA also maintained fast inference, with a real-time factor (RTF) of 0.00423. On the THCHS30-DNS Challenge dataset, CGAN-SECA achieved the highest average PESQ and STOI scores among all evaluated methods, reaching 2.274 and 0.838, respectively, across the −5, 0, and 5 dB conditions. In particular, it produced the best results for both metrics at 0 and 5 dB, demonstrating effective speech enhancement and good generalization across different noise levels and datasets.
SN  - 1932-6203
ER  -