Non Linear AR-Noise Investigated Via a Kalman Filter
The use of autoregressive (AR) models for extracting the information contained in the fluctuating component of a stationary signal represents a well established technique, widely used in the nuclear reactor engineering field. The possibility of treating non stationary fluctuations via a time varying autoregressive (TV AR) model is here considered in the framework of the Kaiman filters theory. Firstly, the vanishing of the covariance matrix of the state vector is prevented by slightly increasing the matrix at each time step. Moreover the forecasting guess of the state vector is improved by the use of a least squares technique. The results of numerical experiments, made on AR models of increasing order up to four are encouraging and suggest the convenience of further investigating the features of the present approach to this non linear noise problem.
KeywordsState Vector Kalman Filter Filter Theory True Coefficient Gaussian White Noise Process
Unable to display preview. Download preview PDF.
- 1.SMORN V, 12 – 16th October 1987, Munich, FRG.Google Scholar
- 4.P. C. Young, A Recursive Approach to Time Series Analysis, Bull. Inst. Math. Appl.. 10: 209 (1974).Google Scholar
- 7.M. Marseguerra and C. M. Porceddu, An Example of Utilization of Kaiman Filters in Time Series Analysis, 19th IMORN, 4 – 16th June 1986, Rome, Italy.Google Scholar