An Adaptive ARMA Spectral Estimator. Part 1.
VIRGINIA POLYTECHNIC INST AND STATE UNIV BLACKSBURG DEPT OF ELECTRICAL ENGINEERING
Pagination or Media Count:
In this paper, a novel procedure for generating an ARMA spectral model of a wide sense stationary time series is developed. The parameters of this model are selected so that they most closely fit a set of Yule-Walker equations which are estimated from a finite set of time series observations. This ARMA modeling method has been found to exhibit a spectral estimation performance which is typically superior to such alternatives as the maximum entropy AR method, classical Fourier procedures MA, and, the Box-Jenkins method ARMA. One of the principal features of this spectral estimation method is the elegant algebraic structure of the linear system of equations which need to be solved when finding the ARMA models parameters. This shift-invariant type structure gives rise to an adaptive algorithmic solution procedure whose computational efficiency is comparable to that achieved by recently developed fast AR algorithmic methods. The details of the adaptive ARMA modeling procedure will be covered in Part 2 of this paper. These dual characteristics of excellent estimation performance and real time adaptive implementation mark this method as being a primary spectral estimation tool. Author
- Statistics and Probability