The problem of estimating the amplitude of a signal is addressed using
higher-order statistics. The probability distribution of the noise is
assumed to be unknown so that the maximum likelihood estimator cannot
be calculated. The estimator is taken as a polynomial of the observat
ion, the coefficients of which are determined so that the estimate is
unbiased with minimum variance. This method generalizes the linear app
roach, and the estimate variance is reduced. The case of linear-quadra
tic estimation is detailed, and numerical examples are presented.