Short-term Rule of Two: Localizing Non-Stationary Noise Events in Swept-Sine Measurements

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

42 Downloads (Pure)

Abstract

Non-stationary noise is notoriously detrimental to room impulse response (RIR) measurements using exponential sine sweeps (ESSs). This work proposes an extension to a method of detecting non-stationary events in ESS measurements that aims at precise localization of the disturbance in the captured signal. The technique uses short-term running cross-correlation as a means to estimate the instantaneous correlation between two sweep signals. Both, the detection threshold and measured correlation, are evaluated on short windows, allowing for accurate analysis of the entire signal. Additional pre-processing steps are applied to improve the robustness of the proposed technique. The approach is tested on various types of simulated and measured non-stationary noise, showing that detection errors did not exceed 23 ms. The method presented in this work increases the robustness of RIR measurements using ESS against non-stationary noise.
Original languageEnglish
Title of host publicationProceedings of AES 154th Convention
PublisherAudio Engineering Society
Number of pages9
ISBN (Electronic)978-1-7138-7778-3
Publication statusPublished - 14 May 2023
MoE publication typeA4 Conference publication
EventAudio Engineering Society Convention - Espoo, Finland
Duration: 13 May 202315 May 2023
Conference number: 154

Conference

ConferenceAudio Engineering Society Convention
Abbreviated titleAES
Country/TerritoryFinland
CityEspoo
Period13/05/202315/05/2023

Keywords

  • room acoustics
  • acoustic measurements
  • impulse response measurement
  • exponentially swept sine
  • sweep measurements
  • cross-correlation
  • non-stationary noise

Fingerprint

Dive into the research topics of 'Short-term Rule of Two: Localizing Non-Stationary Noise Events in Swept-Sine Measurements'. Together they form a unique fingerprint.

Cite this