On the inflation method in adaptive noise-subspace estimator

Abstract

It is shown that Yang and Kaveh's (1987, 1988) inflation method is not necessary to accomplish an adaptive noise-subspace estimator. Without the inflation method, an alternative noise-subspace estimator, involving less computation and simpler parallel implementation, is developed. We also prove that if the initial noise-subspace basis consists of a set of mutually orthonormal vectors, the proposed estimator has a better performance than that with the inflation method both in the mean sense and in the mean-square sense. Computer simulations are provided to substantiate the superiority of the estimator without the inflation method.