On the memoryless robust estimation of the autocorrelation coefficients

Three methods for the estimation of the matrix elements of the correlation matrix of a finite-dependent stationary random sequence are examined: the sample average and two proposed approaches, namely the pseudo maximum likelihood and the pseudo M-estimator. The latter scheme is found as a solution of a Fredholm integral equation. Finally, the existence of a minimax robust design is proved and a suboptimally robust scheme is proposed.<>