| This dissertation is concerned with the processing of multicomponent waveforms to extract information on the location and parameters of partially obscured target signals. A multicomponent waveform is a waveform which is comprised of many copies of other basic waveforms which may have been altered in some systematic way. The components of the waveforms are in general due to some unknown combination of an interfering process and a target process, each of which generate their own multicomponent waveform. The primary examples are drawn from data and models of the underwater scattering and reverberation which result from active sonar transmissions. In this context, the basic waveforms making up the multicomponent waveform are considered to be altered versions of the transmitted waveform. This type of situation also occurs with medical imaging equipment, radar tracking and targeting systems operating in high clutter environments, and communication and telemetry systems which may be faced with operation in multipath transmission channels.; The fundamental problem addressed in this dissertation is one of accurate and fast estimation of the components of a multicomponent waveform when some subset of the components has very closely spaced time delay parameters. This problem is addressed by extending a procedure known as Fast Maximum Likelihood (FML) estimation to the case when very closely spaced components are present. The extension proposed is ‘Fast Maximum Likelihood Estimation with Multiple Signal Initialization’ (FML-MSI), which augments FML with prior information of the time scale parameters of the waveform components and changes the search procedure. The extensions made to FML arise from detailed analyses of the biasing that can occur with the time scale estimates of waveform components when other closely spaced interfering components are present. By implementing a prior probability on the time scale parameter and making an initial constrained search, FML-MSI can estimate components such as these with reduced susceptibility to biased estimates. |
