We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

eess.AS

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Electrical Engineering and Systems Science > Audio and Speech Processing

Title: Data-Efficient Framework for Real-world Multiple Sound Source 2D Localization

Abstract: Deep neural networks have recently led to promising results for the task of multiple sound source localization. Yet, they require a lot of training data to cover a variety of acoustic conditions and microphone array layouts. One can leverage acoustic simulators to inexpensively generate labeled training data. However, models trained on synthetic data tend to perform poorly with real-world recordings due to the domain mismatch. Moreover, learning for different microphone array layouts makes the task more complicated due to the infinite number of possible layouts. We propose to use adversarial learning methods to close the gap between synthetic and real domains. Our novel ensemble-discrimination method significantly improves the localization performance without requiring any label from the real data. Furthermore, we propose a novel explicit transformation layer to be embedded in the localization architecture. It enables the model to be trained with data from specific microphone array layouts while generalizing well to unseen layouts during inference.
Comments: Accepted to IEEE ICASSP 2021. This article supersedes arXiv:2012.05908
Subjects: Audio and Speech Processing (eess.AS); Machine Learning (cs.LG); Sound (cs.SD)
Cite as: arXiv:2012.05533 [eess.AS]
  (or arXiv:2012.05533v3 [eess.AS] for this version)

Submission history

From: Guillaume Le Moing [view email]
[v1] Thu, 10 Dec 2020 09:22:52 GMT (4445kb)
[v2] Tue, 16 Mar 2021 08:07:24 GMT (6175kb)
[v3] Wed, 17 Mar 2021 08:50:36 GMT (6175kb)

Link back to: arXiv, form interface, contact.