A parameter estimation method for model analysis (Q854450)

The paper is devoted to the development of an efficient hybrid algorithm of Structured Nonlinear Total Least Norm (SNTLN) method, and a scheme of separation by parts of parameters. A regularization process for SNTLN method is introduced. The process of parameter estimation separation scheme is formulated and a new algorithm is derived. The proposed algorithm is developed by means of applying both of parameter separation scheme and the regularized structured nonlinear total least squares method. A major benefit of the parameter separation scheme is the reduction of the residual norm steadily to the accurate approximate. It is also not sensitive to the initial choice of the parameters. This scheme guarantees locally convergence of parameters. Numerical simulations on signal processing models are described to illustrate the obtained results and to show that the suggested algorithm is often superior to SNTLN method in the reduction of the residual.