Beamforming in additive α-stable noise using fractional lower order statistics (FLOS)

Abstract

Non-Gaussian statistical signal processing is important when signal or noise deviates from the ideal Gaussian model. Stable distributions are among the most important non-Gaussian models. Minimum noise power, minimum variance distortionless signal response (MNPDR, MVDR) and minimum mean square error (MMSE) beamformers are widely used to estimate the signals in Gaussian noise environments. In this paper, we present a beamforming technique for additive symmetric /spl alpha/-stable (S/spl alpha/S) noise environments. This new technique uses FLOS to formulate a nonlinear cost function which is then minimised to get an optimum weight vector for the array of sensors while the gain in the desired look direction is constrained to be unity. As this nonlinear constrained optimisation problem doesn't have a closed form solution, we use a gradient-based algorithm to estimate the weight vectors. This new algorithm is computationally efficient and can be used with a wide range of stable noise models.