A new approach to adaptive robust joint detection and estimation of discrete-time signals (Q1369093)

The author presents a new approach to the problem of robust joint detection and parameter estimation of signals in \(\varepsilon\)-contaminated noise. Unlike the conventional robust procedures which use linear models for the observations and apply nonlinear transformations to the observed data, the author proposes a nonlinear observation model and linear processing of the nonlinearly observed data. This approach combined with suitable adaptation rules results in algorithms with highly desirable properties, such as fast convergence, good detection capabilities, robustness for short observations. Also, a priori information can be incorporated in a natural way. This new approach overcomes some of the deficiencies of the conventional robust procedures and seems a natural alternative to them for the considered problem.