Generalized regularization in cumulant-based parameter estimation
Jingmin Xin, K. Nagasawa, Akira Sano · 2003
The regularization in third-order cumulant-based parameter estimation is discussed for identification of transfer functions. The SVD-based regularization algorithm to estimate the number of sinusoids in the presence of colored Gaussian noise is given on a basis of fourth-order cumulant. A new MSE-like criterion for determining the optimal multiple adjustable parameters is given to decide the number of sinusoids by comparing these parameters with the singular values of the fourth-order cumulant-based Toeplitz matrix.>