High resolution processing techniques for temporal and spatial signals

Abstract

The first modern generation of signal processing algorithms used Fourier methods. Their theoretical basis was provided by tile convolutions which occur in time-invariant linear systems aad tile important role of sinusoids as eigenfunctions of every such system. A second class of signal analysis and parameter estimation methods use a general linear modcl similar to l, he one used in the statistical literature. P~epresentative methods in this class are the linear predictive and signal subspace methods for parameter estimat. ion, A third class of signal processing methods based on time-frequency analysis of nonstationary signals began with the work of Wigner (1932) and Ville (1948) and has seen extensive theoretical development and the beginning of hardware implementation in the 1978-1988 period. Recent work has Colnbined linear predictive modelling with Wigner-Ville analysis. The near future wiil see the combination of time-frequency distribution (TFD) methods with both eigensystem-based signal analysis and wideband ambiguity function and wideband WVD analysis in which Doppler is treated as a time compression or expansion. This paper provides an updated discussion of high resolution processing techniques for temporal and spatial signal processing, with particular emphasis on methods applicable to underwater acoustic signal processing. The paper consists of three parts: The first part treats direction finding and spectral estimation for stationary random processes. The second part discusses time-frequency analysis for nonstationary signals using both the Wigner-Ville distribution (WVD) and an autoregressive modified WVD. The third part discusses the application of these techniques to radar and sonar imaging.