Robust estimation of signal parameters with dependent data

This paper is concerned with robust estimation of signals of known form in additive non-Gaussian noise which has an autoregressive form of dependency structure. The signals are assumed to be of the linear regression model variety, and there are two main points to the paper. The first is that use of Huber´s [10] ordinary regression M-estimates is not appropriate. The second point is that there do exist proper M-estimates for this problem, and they can be computed by robustifying a two-stage procedure introduced by Durbin [4].