Research On Robust Mode Space Detectors In Complex Shallow Water Environments

This dissertation presents a comprehensive study about the robust detection of low-frequency line spectrum radiated from long-range targets in range-independent shallow-water environments,using received data on a vertical line array(VLA).Considering the source position is unknown and the environment may be uncertain,several robust detectors in presence of different combinations of the unknown model parameters(source position and environmental parameters)are proposed,robust in the sense that the performance only depends on the signal energy impinging on the array and is insensitive to the specific values of the unknown model parameters themselves.The effects of incomplete modal sampling caused by insufficient array apertures on the proposed robust detectors are investigated,and the performance improvement methods are proposed.Compared to the traditional generalized likelihood ratio and Bayesian detectors,the proposed detector avoids their significant performance degradation in certain detection scenarios at a sacrifice of slight performance degradation compared to their performance peaks,and can also save many computational complexities.The main work of this dissertation includes:1.The performances of the commonly-used detectors are investigated and compared,and the performance robustness is introduced.As the motivation of this work,the performances of the commonly-used GLRD,the BD,the energy detector(ED)are investigated and compared in presence of different unknown model parameters,and the reason that leads to each one's performance property is briefly discussed.Based on the performance comparison,the performance robustness expected in this work is introduced.It is the performance property presented by the ED,that is,the performance is only concerned with the signal energy impinging on the array and insensitive to other source or environmental parameters.2.The matched mode space detector(MMSD)and the effective spectrum detector(ESD)when the environment is precisely known is proposed.Utilizing the modal propagation property in shallow-water environments,the robust detector that can provide the robust performance over unknown source position is proposed,that is the MMSD.Due to the utilization of environmental knowledge from the mode space,it obtains a better robust performance than the ED.The effects of incomplete mode sampling on the MMSD is investigated.Via the singular value decomposition of the mode matrix(in the MMSD statistic),the statistic is decomposed into several components.The statistics of each component are derived and the effect of incomplete modal sampling on the performance of each component is investigated.Accordingly,the ESD is proposed to obtain a better robust performance in incompletely mode sampling cases;and the relative signal loss(RSL)is introduced to determine the effective spectrum.3.The robust mode space detector(RMSD)is proposed in uncertain shallow water with complete mode sampling.The relative projection deviation(RPD)is defined to characterize the containing degree between different mode spaces.The simulation results in a modified “genlmis”(general mismatch)waveguide show that due to good modal correlations,the mode spaces have an approximate containing relation that one is nearly contained in another same-or greater-dimensional one.This conclusion is also verified using the collected environmental data from SWell Ex-96 experiment.The statistics of the Mis-MSD are derived,and the relation between the RSL caused by environmental mismatch and the RPD from the true mode space to the mismatched is established.Assuming the mode space also derives from the uncertain environment as the truth does and using the “containing” relation between the mode spaces,the Mis-MSD's performance is further analyzed.From this performance analysis,the MSD using the physically-supported greatest dimensional mode space in uncertain shallow water(RMSD)is proposed to achieve the robustness.4.The greatest dimensional effective MSD(GE-MSD)is proposed in uncertain shallow water with incomplete mode sampling;all the robust detectors presented in this study are summarized in a unified framework.How to obtain the robust performance is investigated by choosing proper subspaces to fulfill the subspace detector(SD).How to get rid of the nuisance modes caused by incomplete modal sampling in the mode matrix / mode space is investigated.The relation between the mode space rejecting the nuisance modes(effective mode space)in different environmental realizations is also analyzed.The distribution of the SD statistic is derived and a sufficient condition for the SD obtaining the robust performance is obtained.From this,the GE-MSD is proposed.It uses the effective mode space corresponding to the greatest dimensional mode space the uncertain shallow-water environment can support for a given source frequency.The simulation results in the “genlmis” waveguide and the real data analysis using SWell Ex-96 experimental data both demonstrate the effectiveness of the GE-MSD.All the robust detectors in this dissertation are summarized in a unified framework: They are the SDs using the subspace satisfying the same sufficient condition,and their differences and connections are also discussed.