Adaptive spectral estimation by the conjugate gradient method

Abstract

This paper proposes an alternative technique for adaptive spectral estimation. The new technique applies the method of conjugate gradient, which is used for iteratively finding the generalized eigenvector corresponding to the minimum generalized eigenvalue of a semidefinite Hermitian matrix, to the adaptive spectral analysis problem. Computer simulations have been performed to compare the new method to existing ones. From the limited examples presented, it is seen that the new method is computationally more efficient at the expense of more core storage. Also, this method is effective for small data records and can implement noise correction to yield unbiased spectral estimates if an estimate of the noise covariance matrix is available. The technique performs well for both narrow-band and wide-band signals.