ON FREQUENCY-WAVE-NUMBER ESTIMATION BY STATE-SPACE REALIZATION

Most existing techniques for frequency-wavenumber estimation require a computationally expensive numerical search over a 2-D parameter space for the maximum or minimum of a function. A method has been proposed using state-space models, which decomposes the 2-D problem into two 1-D problems and provides an analytical solution (i.e., without searching) in each dimension separately. In this paper, we show how the separate estimates obtained by the state-space approach can be grouped into pairs of estimates corresponding to each 2-D signal, and show the relationship with a pairing method given in [17]. In addition, we show how a 2-D forward-backward data matrix can be used to improve the performance of the state-space approach.