Comparing the Sequential Nonlinear least squared Method and Sequential robust M method to estimate the parameters of Two Dimensional sinusoidal signal model:

Abstract

Estimation of the unknown parameters in 2-D sinusoidal signal model can be considered as important and difficult problem. Due to the difficulty to find estimate of all the parameters of this type of models at the same time, we propose sequential non-liner least squares method and sequential robust  M method after their development through the use of sequential  approach in the estimate suggested by Prasad et al to estimate unknown frequencies and amplitudes for the 2-D sinusoidal compounds but depending on Downhill Simplex Algorithm in solving non-linear equations for the purpose of obtaining non-linear parameters estimation which represents frequencies and then use of least squares formula to estimate linear parameters which represents amplitude . solve non-linear equations using Newton –Raphson method in sequential  non-linear least squares method and obtain parameters estimate that represents frequencies and linear parameters which represents amplitude  at the same time, and compared this method with sequential robust  M method when the signal affected by different types of noise including the normal distribution of the error and the heavy-tailed distributions error, numerical simulation are performed to observe the performance of the estimation methods for different sample size, and various level of variance using a statistical measure of mean square error (MSE), we conclude in general that sequential non-linear least squares method is more efficiency compared to others if we follow the normal and logistic distribution of noise, but if the noise follow Cauchy distribution it was a sequential robust M method based on bi-square weight function is the best in the estimation.