Fast estimation of narrowband spectra

A new method is investigated for a fast and consistent estimation of the spectral density of stationary processes which are narrowband. The method exploits the Toeplitz structure of the classical smoothed periodogram estimates and their similarity to circulant forms. The new estimates compare favourably with the classical estimates and, using a fast Fourier transform algorithm, are more rapidly computable than many frequently used estimates when the process is sufficiently narrowband.