Parameter estimation aspects and the role of bandwidth in a notch filter

In this paper, the notch filter model developed for the retrieval of sinusoidal signals in noise is reexamined. A stochastic Gauss-Newton method is developed for adaptively estimating the parameters of the model. Various forms of simplifications (gradient truncation) which assure stability and convergence are then examined. The conditions under which they are applicable is discussed. The bandwidth of the notch filter which depends on a parameter ?? is shown to be crucial for fast and assured convergence.