Sperm Whale Impulse Noise and Incoherent Narrowband CFAR Processing

Impulsive noise interference of a biological origin is common in underwater acoustic environments. Here we report on properties of impulsive noise produced by sperm whales, recorded in deep water off the coast of Manus Island. We analyze the noise as a function of frequency band, and find strong non...

Full description

Bibliographic Details
Main Authors: Bertilone, Derek, Killeen, Damien
Other Authors: DEFENCE SCIENCE AND TECHNOLOGY ORGANISATION EDINBURGH (AUSTRALIA) MARITIME OPERATIONS DIV
Format: Text
Language:English
Published: 2002
Subjects:
Acoustics
Radiofrequency Wave Propagation
*IMPULSE NOISE
*UNDERWATER ACOUSTICS
*WHALES
SIMULATION
DETECTION
PERFORMANCE(ENGINEERING)
PROCESSING
THRESHOLD EFFECTS
THEORY
SIGNAL TO NOISE RATIO
PROCESSING EQUIPMENT
INCOHERENCE
FREQUENCY BANDS
MATHEMATICAL FILTERS
NONLINEAR SYSTEMS
SIGNALS
INTERFERENCE
NOISE
AUSTRALIA
DEEP WATER
SONAR
NARROWBAND
RAYLEIGH WAVES
FADING(ELECTROMAGNETIC WAVES)
AQ F03-06-1400
FOREIGN REPORTS
Sperm whale
Online Access:http://www.dtic.mil/docs/citations/ADA411913
http://oai.dtic.mil/oai/oai?&verb=getRecord&metadataPrefix=html&identifier=ADA411913
Description
Summary:Impulsive noise interference of a biological origin is common in underwater acoustic environments. Here we report on properties of impulsive noise produced by sperm whales, recorded in deep water off the coast of Manus Island. We analyze the noise as a function of frequency band, and find strong non-Gaussian effects in the range 1-6 kHz. Sonar performance can be improved by utilizing nonlinear filtering techniques from non-Gaussian detection theory. We demonstrate this by measuring empirical gains for a constant false-alarm rate (CPAR) incoherent narrowband processor. We investigate both parametric locally optimum and nonparametric filtering of the time-domain data to enhance weak signal detection. Simulated Rayleigh fading signals were inserted into the noise data, and the detection threshold was found to be lowered by about 4 dB over conventional processing, for signal frequencies 1.5 kHz and 4.5 kHz and a false-alarm probability of 0.0001. Performance degradations can potentially occur at large signal-to-noise ratios, but these were mitigated using a simple technique for fusing the non-Gaussian processor with a conventional processor.

Similar Items