Title :
Windows associated with parametric spectral estimators
Author :
Durrani, Tariq S.
Author_Institution :
University of Strathclyde, Glasgow, Scotland, UK
Abstract :
This paper details an analysis for the presence of spectral windows associated with the Maximum Entropy (MESA) Spectral Estimator. By embedding the derivation of the spectral estimator into a constraint optimisation framework a relationship is established between the true spectrum of a second order stationary stochastic sequence and the Maximum Entropy spectrum. This relationship is shown to lead to a spectral window which appears in the reciprocal of the MESA estimator. The structure of the window is based upon a Dirichlet type kernel and depends upon the prediction filter length for a large class of data sequences. Computed results are included which verify the presence of these spectral windows.
Keywords :
Constraint optimization; Cost function; Covariance matrix; Entropy; Filters; Kernel; Lagrangian functions; Maximum likelihood estimation; Spectral analysis; Stochastic processes;
Conference_Titel :
Acoustics, Speech, and Signal Processing, IEEE International Conference on ICASSP '82.
DOI :
10.1109/ICASSP.1982.1171705