Singular Value Spectrum Based Detection and Estimation of Non-StationaryUnderwater Acoustic Signals and Their Statistical Modelling by Approximation

摘要

Underwater acoustic systems detect the presence of objects in the water either bydirectly sensing acoustic energy radiated by the object-passive sonar, or by transmitting an acoustic signal and detecting the reflection from the object-active sonar. In general, the acoustic energy emitted by the source reaches the receiver through different paths: as direct, reflected and/or diffracted waves. The signals may be received at a single sensing point or at multiple sensing points. The processing techniques depend on the number of receiving points and on the number of arrivals at each point. The passive sonar system bases its detection and estimation of target parameters on sounds which emanate from the target itself, including machinery noise, flow noise, transmission from its active sonar, etc. The received signal represents the source signature contaminated by noise. In addition to the direct path, there are multipaths in sound wave propagation. The source signature may include wideband, narrowband and transient components. These signals occur in various multiplicities and have differing spectra, non-planar wavefronts and non-stationary behavior. Noise may be partially correlated with the signal, and high noise coherence may exist between different sensors.