Journal Article

PUBDB-2024-00134

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png StoRM: A Diffusion-Based Stochastic Regeneration Model for Speech Enhancement and Dereverberation

Lemercier, J.-M. ; Richter, J. ; Welker, S.CFEL*DESY* ; Gerkmann, T. (Corresponding author)

2023
IEEE New York, NY

This record in other databases:      

Please use a persistent id in citations: doi:10.1109/TASLP.2023.3294692  doi:10.3204/PUBDB-2024-00134

Report No.: arXiv:2212.11851

Abstract: Diffusion models have shown a great ability at bridging the performance gap between predictive and generative approaches for speech enhancement. We have shown that they may even outperform their predictive counterparts for non-additive corruption types or when they are evaluated on mismatched conditions. However, diffusion models suffer from a high computational burden, mainly as they require to run a neural network for each reverse diffusion step, whereas predictive approaches only require one pass. As diffusion models are generative approaches they may also produce vocalizing and breathing artifacts in adverse conditions. In comparison, in such difficult scenarios, predictive models typically do not produce such artifacts but tend to distort the target speech instead, thereby degrading the speech quality. In this work, we present a stochastic regeneration approach where an estimate given by a predictive model is provided as a guide for further diffusion. We show that the proposed approach uses the predictive model to remove the vocalizing and breathing artifacts while producing very high quality samples thanks to the diffusion model, even in adverse conditions. We further show that this approach enables to use lighter sampling schemes with fewer diffusion steps without sacrificing quality, thus lifting the computational burden by an order of magnitude. Source code and audio examples are available online.

Note: ISSN 2329-9304 not unique: **2 hits**.

---

Contributing Institute(s):

1. FS-CFEL-1 (Group Leader: Henry Chapman) (CFEL-I)
2. FS-CFEL-1 Fachgruppe CFEL-Infrastrukt. (FS-CFEL-1-CFEL)

Research Program(s):

1. 633 - Life Sciences – Building Blocks of Life: Structure and Function (POF4-633) (POF4-633)
2. HIDSS-0002 - DASHH: Data Science in Hamburg - Helmholtz Graduate School for the Structure of Matter (2019_IVF-HIDSS-0002) (2019_IVF-HIDSS-0002)

Experiment(s):

1. No specific instrument

Appears in the scientific report 2023

---

Database coverage:
; ; Current Contents - Engineering, Computing and Technology ; IF < 5 ; JCR ; SCOPUS ; Science Citation Index Expanded ; Thomson Reuters Master Journal List ; Web of Science Core Collection

---

---

Linked articles:

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png Preprint Lemercier, J.-M. ; Richter, J. ; Welker, S. (Corresponding author)CFEL*DESY* ; Gerkmann, T.
StoRM: A Diffusion-Based Stochastic Regeneration Model for Speech Enhancement and Dereverberation
    Files  Fulltext by arXiv.org BibTeX | EndNote: XML, Text | RIS

---