In light of the extensive use of English in daily life, research on improving the quality of noisy English speech plays a crucial role. This paper proposed an enhanced U-Net model for enhancing noisy English speech by incorporating an attention mechanism and optimizing the loss function. The enhancement effect of the method was evaluated using the VoiceBank-DEMAND dataset. The optimized loss function yielded superior enhancement outcomes for noisy English speech. The perceptual evaluation of speech quality (PESQ) was 3.18, the short-time objective intelligibility (STOI) was 0.95, and the CSIG, CBAK, and COVL were 4.41, 3.65, and 3.83, respectively. These results outperformed other U-Net improved models and existing models. The findings validate the efficacy of the proposed approach for enhancing noisy English speech, thereby demonstrating its practical applicability.
SubbaiahPVDeepthiH. A study to analyze enhancement techniques on sound quality for bone conduction and air conduction speech processing. Scalable Computing2020; 21(1): 57–62.
2.
SunXLiRLiT, et al.Single channel speech enhancement Algorithm based on BLSTM-DNN bidirectional optimized hybrid model. IOP Conf Ser Mater Sci Eng2020; 719(1): 1–6.
3.
JeevaMPANagarajanTVijayalakshmiP. Adaptive multi-band filter structure-based far-end speech enhancement. IET Signal Process2020; 14(5): 288–299.
4.
TyagiSVárkonyi-KóczyARSzénásiS. A comprehensive investigation into the noise reduction techniques for speech. In: 2023 IEEE 21st World Symposium on Applied Machine Intelligence and Informatics (SAMI), Herl'any, Slovakia, 2023, pp. 000207–000212. DOI: 10.1109/SAMI58000.2023.10044486.
5.
DowerahSSerizelRJouvetD, et al.Joint optimization of diffusion probabilistic-based multichannel speech enhancement with far-field speaker verification. In: 2022 IEEE spoken Language technology workshop (SLT), Doha, Qatar, 2023, pp. 428–435. DOI: 10.1109/SLT54892.2023.10022350.
6.
PengWZhenSChenX, et al.Study on convolutional recurrent neural networks for speech enhancement in fiber-optic microphones. J Phys: Conf Ser2022; 2246(1): 012084.
7.
ShimadaKBandoYMimuraM, et al.Unsupervised speech enhancement based on Multichannel NMF-informed beamforming for noise-robust automatic speech recognition. IEEE/ACM Trans Audio Speech Lang Process2019; 27(5): 960–971.
8.
XiangYHøjvangJLRasmussenMH, et al.A bayesian permutation training deep representation learning method for speech enhancement with variational autoencoder. In: ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Singapore, Singapore, 2022, pp. 381–385. DOI: 10.1109/ICASSP43922.2022.9747036.
9.
WalczynaTPiotrowskiZ. Wave-U-Net speech denoising. In: IBIMA Conference on Artificial intelligence and Machine Learning, Madrid, Spain, 2024, pp. 52–57. DOI: 10.1007/978-3-031-62843-6_5.
10.
JannuCVanambathinaS. Shuffle attention U-Net for speech enhancement in time domain. Int J Image Graph2023; 24(4): 1–21.
11.
GuimaresHRNaganoHSilvaDW. Monaural speech enhancement through deep Wave-U-Net. Expert Syst Appl2020; 158: 113582.
12.
Al-BarhanHAElyassSMSaeedTR, et al.Modified speech separation deep learning network based on hamming window. IOP Conf Ser Mater Sci Eng2021; 1076(1): 012059.
13.
ZhangJLiJWangS, et al.An improved unscale S-Transform in frequency domain. IEEE Geosci Remote Sens Lett2023; 20: 1–5.
14.
YinHShaoY. CFU-Net: a coarse–fine U-Net with multilevel attention for medical image segmentation. IEEE Trans Instrum Meas2023; 72: 1–12.
15.
ZangYYuZXuK, et al.Fiber communication receiver models based on the multi-head attention mechanism. Chin Opt Lett2023; 21(3): 1–34.
16.
Valentini-BotinhaoCWangXTakakiS, et al.Investigating RNN-Based speech enhancement methods for noise-robust text-to-speech. In: 9th ISCA Speech Synthesis Workshop, 2016, pp. 159–165. DOI: 10.21437/SSW.2016-24.
17.
VeauxCYamagishiJKingS. The voice bank corpus: design, collection and data analysis of a large regional accent speech database. In: 2013 International Conference Oriental COCOSDA held jointly with 2013 Conference on Asian Spoken Language Research and Evaluation (O-COCOSDA/CASLRE), Gurgaon, India, 2013, pp. 1–4. DOI: 10.1109/ICSDA.2013.6709856.
18.
ThiemannJItoNVincentE. The diverse environments multi-channel acoustic noise database: a database of multichannel environmental noise recordings. J Acoust Soc Am2013; 133(5): 3591.
19.
RixAWBeerendsJGHollierMP, et al.Perceptual evaluation of speech quality (PESQ)-a new method for speech quality assessment of telephone networks and codecs. In: 2001 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings (Cat. No.01CH37221), Salt Lake City, UT, USA, 2001, pp. 749–752. DOI: 10.1109/ICASSP.2001.941023.
20.
TaalCHHendriksRCHeusdensR, et al.An algorithm for intelligibility prediction of time–frequency weighted noisy speech. IEEE Trans Audio Speech Lang Process2011; 19(7): 2125–2136.
21.
HuYLoizouPC. Evaluation of objective quality measures for speech enhancement. IEEE Trans Audio Speech Lang Process2008; 16(1): 229–238.
22.
KohlSAARomera-ParedesBMeyerC, et al.A probabilistic U-net for segmentation of ambiguous images. In: Proceedings of the 32nd International Conference on Neural Information Processing Systems, 2018, pp. 6965–6975. DOI: 10.48550/arXiv.1806.05034.
23.
ChoiHSKimJHHuhJ, et al. Phase-aware speech enhancement with deep complex U-Net. arXiv preprint. 2019. DOI: 10.48550/arXiv.1903.03107.
24.
DefossezASynnaeveGAdiY. Real time speech enhancement in the waveform domain. arXiv preprint. 2020. DOI: 10.48550/arXiv.2006.12847.
25.
LimJSAOppenheimAV. All-pole modeling of degraded speech. IEEE Trans Acoust1978; 26(3): 197–210.
TanKWangD. Complex spectral mapping with a convolutional recurrent network for monaural speech enhancement. In: ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Brighton, UK, 2019, pp. 6865–6869. DOI: 10.1109/ICASSP.2019.8682834.
28.
TanKWangD. Learning complex spectral mapping with gated convolutional recurrent networks for monaural speech enhancement. IEEE/ACM Trans Audio Speech Lang Process2020; 28: 380–390.
29.
WangKHeBZhuWP. TSTNN: two-stage transformer based neural network for speech enhancement in the time domain. In: ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Toronto, ON, Canada, 2021, pp. 7098–7102. DOI: 10.1109/ICASSP39728.2021.9413740.
