Coarse-to-fine Optimization for Speech Enhancement

被引:4
|
作者
Yao, Jian [1 ]
Al-Dahle, Ahmad [1 ]
机构
[1] Apple Inc, Cupertino, CA 95014 USA
来源
关键词
speech enhancement; coarse-to-fine; deep learning; generative model; discriminative model; dynamic perceptual loss;
D O I
10.21437/Interspeech.2019-2792
中图分类号
R36 [病理学]; R76 [耳鼻咽喉科学];
学科分类号
100104 ; 100213 ;
摘要
In this paper, we propose the coarse-to-fine optimization for the task of speech enhancement. Cosine similarity loss [1] has proven to be an effective metric to measure similarity of speech signals. However, due to the large variance of the enhanced speech with even the same cosine similarity loss in high dimensional space, a deep neural network learnt with this loss might not be able to predict enhanced speech with good quality. Our coarse-to-fine strategy optimizes the cosine similarity loss for different granularities so that more constraints are added to the prediction from high dimension to relatively low dimension. In this way, the enhanced speech will better resemble the clean speech. Experimental results show the effectiveness of our proposed coarse-to-fine optimization in both discriminative models and generative models. Moreover, we apply the coarse-to-fine strategy to the adversarial loss in generative adversarial network (GAN) and propose dynamic perceptual loss, which dynamically computes the adversarial loss from coarse resolution to fine resolution. Dynamic perceptual loss further improves the accuracy and achieves state-of-the-art results compared with other generative models.
引用
收藏
页码:2743 / 2747
页数:5
相关论文
共 50 条
  • [41] Coarse-to-fine underwater image enhancement based on multi-level wavelet transform
    Yuan G.
    Yang G.
    Wang J.
    Liu H.
    Wang W.
    Guangxue Jingmi Gongcheng/Optics and Precision Engineering, 2022, 30 (22): : 2939 - 2951
  • [42] Coarse-to-Fine Low-Light Image Enhancement With Light Restoration and Color Refinement
    Wu, Xu
    Lai, Zhihui
    Yu, Shiqi
    Zhou, Jie
    Liang, Zhuoqian
    Shen, Linlin
    IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTATIONAL INTELLIGENCE, 2024, 8 (01): : 591 - 603
  • [43] Coarse-to-fine underwater image enhancement with lightweight CNN and attention-based refinement
    Khandouzi, Ali
    Ezoji, Mehdi
    JOURNAL OF VISUAL COMMUNICATION AND IMAGE REPRESENTATION, 2024, 99
  • [44] Coarse-to-Fine Spatio-Temporal Information Fusion for Compressed Video Quality Enhancement
    Luo, Dengyan
    Ye, Mao
    Li, Shuai
    Li, Xue
    IEEE SIGNAL PROCESSING LETTERS, 2022, 29 : 543 - 547
  • [45] Competing fronts for coarse-to-fine surface reconstruction
    Sharf, Andrei
    Lewiner, Thomas
    Shamir, Ariel
    Kobbelt, Leif
    Cohen-Or, Daniel
    COMPUTER GRAPHICS FORUM, 2006, 25 (03) : 389 - 398
  • [46] A coarse-to-fine framework to efficiently thwart plagiarism
    Zhang, Haijun
    Chow, Tommy W. S.
    PATTERN RECOGNITION, 2011, 44 (02) : 471 - 487
  • [47] A Coarse-to-Fine Framework for Point Voxel Transformer
    Bai, Zhuhua
    Meng, Fantong
    Li, Weiqing
    Kang, Renke
    Yang, Guolin
    Dong, Zhigang
    PROCEEDINGS OF THE 2024 27 TH INTERNATIONAL CONFERENCE ON COMPUTER SUPPORTED COOPERATIVE WORK IN DESIGN, CSCWD 2024, 2024, : 205 - 211
  • [48] Multiscale Coarse-to-Fine Guided Screenshot Demoireing
    Nguyen, Duong Hai
    Lee, Se-Ho
    Lee, Chul
    IEEE SIGNAL PROCESSING LETTERS, 2023, 30 : 898 - 902
  • [49] CNN with coarse-to-fine layer for hierarchical classification
    Fu, Ruigang
    Li, Biao
    Gao, Yinghui
    Wang, Ping
    IET COMPUTER VISION, 2018, 12 (06) : 892 - 899
  • [50] Robust Weighted Coarse-to-Fine Sparse Tracking
    Zhong, Boxuan
    Chen, Zijing
    You, Xinge
    Li, Luoging
    Xie, Yunliang
    Yu, Shujian
    2014 INTERNATIONAL CONFERENCE ON SECURITY, PATTERN ANALYSIS, AND CYBERNETICS (SPAC), 2014, : 7 - 14