Acoustic signature identification using distributed diffusion adaptive networks

Sayed Mostafa Taheri*, Hamed Nosrati

*Corresponding author for this work

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

Abstract

In this paper, we propose using distributed diffusion adaptive networks for acoustic signature identification, as a time-varying autoregressive (TVAR) stochastic model. A distributed adaptive sensor network considers spatio-temporal challenges simultaneously. To analyze diffusion networks under TVAR modeling problem circumstances, we investigate and elaborate on their performance under non-stationary conditions. Different versions of diffusion networks are then theoretically compared under the problem conditions. Furthermore, their superiority to single point observations is shown. Finally, the proposed algorithms are implemented on a raw and real sensor network dataset recorded from moving vehicles. The experimental results well support the theoretical findings, and demonstrate the excellence and efficacy of distributed diffusion adaptive networks for this case.

Original languageEnglish
Title of host publication2014 9th International Symposium on Communication Systems, Networks and Digital Signal Processing, CSNDSP 2014
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages943-948
Number of pages6
ISBN (Print)9781479925810
DOIs
Publication statusPublished - 1 Jan 2014
Externally publishedYes
Event2014 9th International Symposium on Communication Systems, Networks and Digital Signal Processing, CSNDSP 2014 - Manchester, United Kingdom
Duration: 23 Jul 201425 Jul 2014

Conference

Conference2014 9th International Symposium on Communication Systems, Networks and Digital Signal Processing, CSNDSP 2014
CountryUnited Kingdom
CityManchester
Period23/07/1425/07/14

Keywords

  • Acoustic signature identification
  • Adaptive networks
  • Diffusion LMS
  • Distributed estimation
  • TVAR modelling

Fingerprint

Dive into the research topics of 'Acoustic signature identification using distributed diffusion adaptive networks'. Together they form a unique fingerprint.

Cite this