A MMSE Filter for Range Sidelobe Reduction

In many cases, one needs to estimate a target that is obscured by the sidelobes of a very large, adjacent scatterer. For example, the Earth's surface can ruin nearsurface rain-rate measurements obtained from an airborne platform, or it can prevent a ground-penetrating radar from detecting land mines located at shallow depth. In these cases, the range sidelobes of the ground are larger than the targets near the surface. The range sidelobes cannot be optimally controlled because target information is not used. However, in most situations the location and magnitude of such a large scatterer can be predicted. In this paper, a minimum meansquared error (MMSE) estimator is proposed to improve the range sidelobe problem. The MMSE estimator uses information about the scenario to find the optimum processing filter. Simulated results are provided to demonstrate the effectiveness of this process. INTRODUCTION It is often desired that a radar detect a target located in close proximity to a large, specular scatterer. For example, downward-looking radars, as in [1], may need to detect targets just above the Earth, or ground-penetrating radars may need to detect targets buried just beneath the air-soil interface. Typically, the approach taken to this problem is to use as much bandwidth as possible [2-4] and to apply a window function to the frequency spectrum of the measurements. More signal bandwidth increases the effective number of range bins between the target and specular while windowing sacrifices some of that bandwidth for reduced sidelobes. Although increased bandwidth certainly increases the resolution between two targets, a window function is rarely the optimum weighting function of choice for a weakly scattering target located in the first few sidelobes of a large scatterer. If a priori knowledge of the statistical properties of both targets is available, this information can be used to calculate the optimum weighting function that should be applied to the measurement spectrum. This optimum weighting function is calculated using the linear minimum mean-squared error (MMSE) estimator. Furthermore, the weighting function can be applied as part of the processing filter, rather than as weights on the data. In this way, the uniquely optimum weights for each range cell can be calculated and applied independently. In this paper, the advantage of the MMSE estimator is confirmed through calculation and presentation of ambiguity functions. For two targets at different ranges from specular, the MMSE filter is calculated and correlated with the received spectrum as a function of range. The resulting ambiguity functions demonstrate the advantage of using MMSE estimation.