Local modeling and robust estimation for high-resolution direction finding

Many signal processing applications are concerned with estimating signal parameters with high accuracy from sensor array measurements. To obtain accurate location estimates, knowledge of the array response is required. Most estimation techniques are sensitive to perturbations in the array model and array calibration is often necessary. The authors discuss the issue of generating the array response parameterization from a finite number of noise-contaminated calibration vectors. An estimation technique based on local parametric modeling of the array response is proposed. The potential improvement using the suggested scheme rather than an ideal array model is demonstrated on real data collected from a full-scale hydroacoustic array. It is shown that, by applying the proposed technique together with a standard direction-of-arrival estimation procedure, a significant reduction in estimation bias can be obtained.<<ETX>>