Publication detail

Identification of regularization parameter for NAH by comparison of results of different NAH calculation methods

HAVRÁNEK, Z. KLUSÁČEK, S.

Original Title

Identification of regularization parameter for NAH by comparison of results of different NAH calculation methods

English Title

Identification of regularization parameter for NAH by comparison of results of different NAH calculation methods

Type

conference paper

Language

en

Original Abstract

The planar near-field acoustical holography (NAH) based on spatial transformations and involves recursive Wiener filtering performs backward calculation of sound field towards source surface. This inverse calculation causes strong amplification of evanescent waves and undesirable errors incorporated in measurement data, e.g. transducers mismatch and environmental noise. Also other different planar NAH algorithms, avoiding spatial transformations, e.g. statistically optimized near-field acoustical holography (SONAH), struggle with errors in input (measured) data. Consequently, some amount of regularization in NAH calculation is needed. Estimation of proper amount of regularization for real world input data is quite difficult task and many methods try to solve it, e.g. Tikhonov regularization with L-curve analysis or GCV. This paper describes and evaluates other method for estimating regularization parameter. The method uses results from two NAH algorithms with different calculation procedure (e.g. SONAH & classical NAH), compares resulting sound maps and in recursive calculation tries to minimize mean square error of small region in the centre of the calculated planar surface by modification of regularization parameter in both methods. Results from simulated measurements with various true SNR in measured data (different transducer mismatch) show quite high successfulness of estimation of proper regularization to minimize NAH projection calculation error.

English abstract

The planar near-field acoustical holography (NAH) based on spatial transformations and involves recursive Wiener filtering performs backward calculation of sound field towards source surface. This inverse calculation causes strong amplification of evanescent waves and undesirable errors incorporated in measurement data, e.g. transducers mismatch and environmental noise. Also other different planar NAH algorithms, avoiding spatial transformations, e.g. statistically optimized near-field acoustical holography (SONAH), struggle with errors in input (measured) data. Consequently, some amount of regularization in NAH calculation is needed. Estimation of proper amount of regularization for real world input data is quite difficult task and many methods try to solve it, e.g. Tikhonov regularization with L-curve analysis or GCV. This paper describes and evaluates other method for estimating regularization parameter. The method uses results from two NAH algorithms with different calculation procedure (e.g. SONAH & classical NAH), compares resulting sound maps and in recursive calculation tries to minimize mean square error of small region in the centre of the calculated planar surface by modification of regularization parameter in both methods. Results from simulated measurements with various true SNR in measured data (different transducer mismatch) show quite high successfulness of estimation of proper regularization to minimize NAH projection calculation error.

Keywords

acoustic holography, regularization, acoustic imagining, sound sources, holography algorithms

RIV year

2007

Released

28.08.2007

Publisher

Turkish Acoustical Society

Location

Istanbul

ISBN

978-1-60560-385-8

Book

INTER-NOISE 2007 Proceedings on CD-ROM

Edition number

1

Pages from

1

Pages to

9

Pages count

9

Documents

BibTex


@inproceedings{BUT23165,
  author="Zdeněk {Havránek} and Stanislav {Klusáček}",
  title="Identification of regularization parameter for NAH by comparison of results of different NAH calculation methods",
  annote="The planar near-field acoustical holography (NAH) based on spatial transformations and involves recursive Wiener filtering performs backward calculation of sound field towards source surface. This inverse calculation causes strong amplification of evanescent waves and undesirable errors incorporated in measurement data, e.g. transducers mismatch and environmental noise. Also other different planar NAH algorithms, avoiding spatial transformations, e.g. statistically optimized near-field acoustical holography (SONAH), struggle with errors in input (measured) data. Consequently, some amount of regularization in NAH calculation is needed. Estimation of proper amount of regularization for real world input data is quite difficult task and many methods try to solve it, e.g. Tikhonov regularization with L-curve analysis or GCV. This paper describes and evaluates other method for estimating regularization parameter. The method uses results from two NAH algorithms with different calculation procedure (e.g. SONAH & classical NAH), compares resulting sound maps and in recursive calculation tries to minimize mean square error of small region in the centre of the calculated planar surface by modification of regularization parameter in both methods. Results from simulated measurements with various true SNR in measured data (different transducer mismatch) show quite high successfulness of estimation of proper regularization to minimize NAH projection calculation error.",
  address="Turkish Acoustical Society",
  booktitle="INTER-NOISE 2007 Proceedings on CD-ROM",
  chapter="23165",
  howpublished="electronic, physical medium",
  institution="Turkish Acoustical Society",
  year="2007",
  month="august",
  pages="1--9",
  publisher="Turkish Acoustical Society",
  type="conference paper"
}