Directional NMF for joint source localization and separation

Johannes Traa, Paris Smaragdis, Noah D. Stein, David Wingate

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

We propose an unsupervised method for simultaneously localizing and separating speech signals by factorizing a non-negative matrix of Steered Response Power (SRP) measurements. We use a probabilistic interpretation of the SRP function to compute a wideband SRP matrix. Non-negative Matrix Factorization (NMF) is used to decompose it into three terms that describe (1) the source distributions over spatial directions, (2) the overall source activations, and (3) the source activations over the time-frequency (TF) plane. The first term indicates the sources' Directions-of-Arrival (DOA) and the latter two terms provide TF weights for separating the sources. Experiments show that this joint approach out-performs a sequential SRP localization + beamforming method.

Original languageEnglish (US)
Title of host publication2015 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2015
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781479974504
DOIs
StatePublished - Nov 24 2015
EventIEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2015 - New Paltz, United States
Duration: Oct 18 2015Oct 21 2015

Publication series

Name2015 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2015

Other

OtherIEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2015
Country/TerritoryUnited States
CityNew Paltz
Period10/18/1510/21/15

Keywords

  • beamforming
  • nonnegative matrix factorization
  • source localization
  • steered response power

ASJC Scopus subject areas

  • Computer Science Applications
  • Signal Processing
  • Media Technology

Fingerprint

Dive into the research topics of 'Directional NMF for joint source localization and separation'. Together they form a unique fingerprint.

Cite this