A new robust method for 2-D sinusoidal frequency estimation

Abstract

Summary form only given, as follows. A new one-dimensional (1-D) frequency estimation method for two-dimensional (2-D) sinusoidal signals is proposed. The received observations are assumed to be a sum of 2-D sinusoidal signals and additive white noise process. The distribution of the noise process is not necessarily Gaussian. Among the family of robust estimates, the M-estimate which is obtained by the minimization of a nonquadratic function of normalized residuals, is considered. It is shown that the new frequency estimates perform as well as a 2-D autoregressive (AR) model based method. Further, in the new estimation method the variance of resulting estimate is guaranteed to be O(N/sup -2/). Numerical simulation studies confirming the performance of the new estimation method are presented.<>