Source Separation with Deep Generative Priors

被引:0
|
作者
Jayaram, Vivek [1 ]
Thickstun, John [1 ]
机构
[1] Univ Washington, Paul G Allen Sch Comp Sci & Engn, Seattle, WA 98195 USA
基金
美国国家科学基金会;
关键词
BLIND SOURCE SEPARATION;
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Despite substantial progress in signal source separation, results for richly structured data continue to contain perceptible artifacts. In contrast, recent deep generative models can produce authentic samples in a variety of domains that are indistinguishable from samples of the data distribution. This paper introduces a Bayesian approach to source separation that uses generative models as priors over the components of a mixture of sources, and noise-annealed Langevin dynamics to sample from the posterior distribution of sources given a mixture. This decouples the source separation problem from generative modeling, enabling us to directly use cutting-edge generative models as priors. The method achieves state-of-the-art performance for MNIST digit separation. We introduce new methodology for evaluating separation quality on richer datasets, providing quantitative evaluation of separation results on CIFAR-10. We also provide qualitative results on LSUN.
引用
收藏
页数:12
相关论文
共 50 条
  • [41] Global Guarantees for Blind Demodulation with Generative Priors
    Hand, Paul
    Joshi, Babhru
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 32 (NIPS 2019), 2019, 32
  • [42] The Generalized Lasso with Nonlinear Observations and Generative Priors
    Liu, Zhaoqiang
    Scarlett, Jonathan
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 33, NEURIPS 2020, 2020, 33
  • [43] DETERMINED AUDIO SOURCE SEPARATION WITH MULTICHANNEL STAR GENERATIVE ADVERSARIAL NETWORK
    Li, Li
    Kameoka, Hirokazu
    Makino, Shoji
    PROCEEDINGS OF THE 2020 IEEE 30TH INTERNATIONAL WORKSHOP ON MACHINE LEARNING FOR SIGNAL PROCESSING (MLSP), 2020,
  • [44] Initialization for NMF-Based Audio Source Separation Using Priors on Encoding Vectors
    Jaeuk Byun
    Jong Won Shin
    中国通信, 2019, 16 (09) : 177 - 186
  • [45] Initialization for NMF-Based Audio Source Separation Using Priors on Encoding Vectors
    Byun, Jacuk
    Shin, Jong Won
    CHINA COMMUNICATIONS, 2019, 16 (09) : 177 - 186
  • [46] Bayesian Source Separation of Linear and Linear-quadratic Mixtures Using Truncated Priors
    Leonardo Tomazeli Duarte
    Christian Jutten
    Saïd Moussaoui
    Journal of Signal Processing Systems, 2011, 65 : 311 - 323
  • [47] Bayesian Source Separation of Linear and Linear-quadratic Mixtures Using Truncated Priors
    Duarte, Leonardo Tomazeli
    Jutten, Christian
    Moussaoui, Said
    JOURNAL OF SIGNAL PROCESSING SYSTEMS FOR SIGNAL IMAGE AND VIDEO TECHNOLOGY, 2011, 65 (03): : 311 - 323
  • [48] MUSIC SOURCE SEPARATION WITH DEEP EQUILIBRIUM MODELS
    Koyama, Yuichiro
    Murata, Naoki
    Uhlich, Stefan
    Fabbro, Giorgio
    Takahashi, Shusuke
    Mitsufuji, Yuki
    2022 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP), 2022, : 296 - 300
  • [49] DEEP VARIATIONAL GENERATIVE MODELS FOR AUDIO-VISUAL SPEECH SEPARATION
    Viet-Nhat Nguyen
    Sadeghi, Mostafa
    Ricci, Elisa
    Alameda-Pineda, Xavier
    2021 IEEE 31ST INTERNATIONAL WORKSHOP ON MACHINE LEARNING FOR SIGNAL PROCESSING (MLSP), 2021,
  • [50] Deep Quasi-Periodic Priors: Signal Separation in Wearable Systems with Limited Data
    Saffarpour, Mahya
    Vali, Kourosh
    Qian, Weitai
    Kasap, Begum
    Farmer, Diana L.
    Wang, Aijun
    Ghiasi, Soheil
    2024 DESIGN, AUTOMATION & TEST IN EUROPE CONFERENCE & EXHIBITION, DATE, 2024,