Bayesian array signal processing in additive generalized Gaussian noise

We present a Bayesian approach for DOA and frequency estimation of narrow band signals in additive generalized Gaussian noise. Using Bayesian techniques, the posterior probability densities for DOA (direction of arrival) and frequency parameters are derived from the signal and noise models. These posterior probabilities are then used in the Metropolis-Hastings (M-H) algorithm to derive the samples for the DOA and frequency parameters. The performances of our algorithms are studied by plotting the MSEs (mean square errors) of the parameters for various SNRs. The MSEs of the parameters are compared with the CRLBs (Cramer Rao lower bound) for the generalized Gaussian models.

1. Introduction

Sensor array processing has found important applications in many areas such as radar, sonar, communications and seismic explorations. Determination of DOAs and the frequencies of the transmitted signals are two of the main problems in sensor array processing. Various methods have been proposed for estimating DOA and associated parameters for multiple plane-waves signals incident on array of sensors. These include subspace-based methods [5] and maximum likelihood techniques [4]. The signal processing literature has traditionally been dominated by Gaussian noise model assumptions. However, many classes of noise encountered in the real-world such as underwater acoustic noise, low frequency electro-magnetic disturbances [3] and atmospheric noise, exhibit outliers that will not fit into a Gaussian noise model.