PASSIVE TARGET TRACKING USING MAXIMUM-LIKELIHOOD-ESTIMATION

Estimation of target trajectory from passive sonar bearings and freque ncy measurements in the presence of multivariate normally distributed noise, with unknown inhomogeneous general covariance, is modeled as a nonlinear multiresponse parameter estimation problem. It is shown that maximum Likelihood estimation in this case is identical to optimizing a determinant criterion which has a concise form and contains no elem ents of unknown covariance matrix. A Gauss-Newton type algorithm, usin g only the first-order derivatives of the model function and a new con vergence criterion, is presented to implement such estimation. The sim ulation results demonstrate that performance of the maximum likelihood estimation method with the above noise model is superior to that with the traditional noise assumption.