Improving the performance of AR estimation of near-field signal sources by decimation

The problem of locating signal sources in the near-field of a linear sensory array is considered. To locate near-field signal sources, the overall array is divided into a number of smaller subarrays. The direction-of-arrival (DOA) of the signal for each subarray is then determined from an autoregressive (AR) modeling of the subarray data. It was previously shown that in AR parameter estimation. decimation will increase the capability of the power spectral density in resolving closely spaced sinusoids. It is shown that decimation, in addition to providing higher resolution (ability to separate two nearby peaks), also decreases the DOA perturbations caused by perturbations in the pole location estimates. Effects of decimation in locating near-field signal sources are presented with an example of simulated fields for a combination of near- and far-field sources.<>