Speaker localization using the direct-path dominance test for arbitrary arrays

Hanan Beit-On, Boaz Rafaely

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Speaker localization in reverberant environments is an important and timely challenge in audio signal processing, with new methods recently developed to meet this challenge. One of these methods is the direct-path dominance (DPD) test showing robustness to both reverberation and noise. However, this method has been developed for processing in the spherical harmonics domain and is restricted to microphone arrays with a spherical configuration. In this paper, we propose a new method for extending the DPD test approach to arrays with arbitrary configuration, by removing the frequency dependence of the steering matrices using local focusing transformations to support frequency-smoothing and robustness to reverberation. A simulation study for evaluating the performance of the method for a binaural array is presented, validating the theoretical developments.

Original languageAmerican English
Title of host publication2018 IEEE International Conference on the Science of Electrical Engineering in Israel, ICSEE 2018
ISBN (Electronic)9781538663783
DOIs
StatePublished - 2 Jul 2018
Event2018 IEEE International Conference on the Science of Electrical Engineering in Israel, ICSEE 2018 - Eilat, Israel
Duration: 12 Dec 201814 Dec 2018

Publication series

Name2018 IEEE International Conference on the Science of Electrical Engineering in Israel, ICSEE 2018

Conference

Conference2018 IEEE International Conference on the Science of Electrical Engineering in Israel, ICSEE 2018
Country/TerritoryIsrael
CityEilat
Period12/12/1814/12/18

Keywords

  • Direction-of-Arrival estimation
  • direct-path
  • room reverberation
  • speaker localization

All Science Journal Classification (ASJC) codes

  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'Speaker localization using the direct-path dominance test for arbitrary arrays'. Together they form a unique fingerprint.

Cite this