Model-based and matched-filterbank signal analysis

Sammanfattning: The dissertation deals with model-based and matched-filterbank signal analysis. The matched-filterbank (MAFI) spectral estimation approach is introduced, and it is shown that both the amplitude spectrum Capon (ASC) and the amplitude and phase estimation (APES) spectral estimators can be expressed as MAFI spectral estimators. A combined estimation procedure for data with mixed spectrum is introduced, as well as ASC and APES implementations for real-valued data. Computationally efficient implementations of the 2-D power spectrum Capon (PSC) and the 1-D and 2-D ASC are proposed.An asymptotic Cramér-Rao bound for line-spectra estimation is derived. It is shown that the non-linear least squares method (NLSM) will asymptotically achieve the same statistical performance as the maximum likelihood method (MLM) even in the colored noise case. Sufficient conditions for identifiability are derived for known and unknown waveforms received through a multipath channel. Statistically efficient subspace-based estimators for the estimation of the time-delay and Doppler parameters are presented.