Abstract
Underwater active sound detection systems face challenges in ocean environments due to ambient noise power levels, reverberation inherent with rough and moving boundaries, and the volume scattering associated with the inhomogeneous media. In addition, ocean waveguides spread acoustic energy in angle and delay leading to images of the target body appearing at the receiver array via the set of eigen rays associated with the refractive waveguide. Such multipath angle and delay spreading can confound more conventional detection schemes. In this thesis, an axiomatic probabilistic approach, in concert with a ray theoretic modeling of the refractive waveguide is employed to construct an environmentally informed Bayes factor (BF) detector for high frequency monostatic active sonar operating in ocean waveguides. A Laplace approximation is employed to marginalize uncertainty in target body depth under the case of the composite alternative. Likewise, a multivariate Gaussian model suitable for high frequency reverberation allows for closed form marginalization under the composite null hypothesis of no target present. It is found that the BF statistic takes the form of a time varying quadratic form that incorporates relevant waveguide information about the ocean environment as well as constraints on the target body depth. The structure of this quadratic form is illuminated and discussed and shown to be a time varying covariance test associated with the target body’s angle delay response at the receiver array as a function of range. Case studies are employed on isovelocity, Pekeris, Munk, and constant gradient ocean profiles to develop and compare the time varying angle delay structure and reveal important features of the waveguide that provide the most information regarding the scattering body’s presence. The key result is the BF extraction of depth invariant modes that provide significant leverage for detection.