Open AccessOperational monitoring of acoustic sensor networks
Author(s) -
V. A. Boltenkov,
G. Kh. M. Al-Jasri
Publication year - 2015
Publication title -
tehnologiâ i konstruirovanie v èlektronnoj apparature
Language(s) - English
Resource type - Journals
eISSN - 2309-9992
pISSN - 2225-5818
DOI - 10.15222/tkea2015.2-3.39
Subject(s) - multilateration , microphone , computer science , signal (programming language) , acoustics , noise (video) , acoustic source localization , cepstrum , cross correlation , speech recognition , acoustic wave , telecommunications , artificial intelligence , node (physics) , mathematics , sound pressure , statistics , physics , image (mathematics) , programming language
Acoustic sensor networks (ASN) are widely used to monitor water leaks in the power generating systems. Since the ASN are used in harsh climatic conditions the failures of microphone elements of ASN are inevitable. That's why the failure detection of ASN elements is a problem of current interest. Two techniques of operational monitoring ASN are developed. Both of them are based on the placement of the test sound source within a network. The signal processing for ASN sensors had to detect the failed element. Techniques are based time difference of arrival (TDOA) estimating at the each pair of ASN elements. TDOA estimates as argmaximum of cross-correlation function (CCF) for signals on each microphone sensors pair. The M-sequence phase-shift keyed signal is applied as a test acoustic signal to ensure high accuracy of the CCF maximum estimation at low signal/noise ratio (SNR). The first technique is based on the isolation principle for TDOA sum at three points. It require to locate the test sound source in the far field. This is not always possible due to technological reasons. For the second proposed technique test sound source can be located near the ASN. It is based on a system of hyperbolic equations solving for each of the four elements of the ASN. Both techniques has been tested in the computer imitation experiment. It was found that for the SNR to –5 dB both techniques show unmistakable indicators of control quality. The second method requires significantly more time control.