Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
NCTM or The Math Forum.



System identification TBx
Posted:
Jul 14, 1997 2:17 PM


I use the system identification toolbox in order to find ARMA models for random processes.
One I know the number of poles and zeros of my model, I use the state description of the model because I put it in a Kalman filter.
I understand how the matrix A (dynamic matrix),K (noise coupling matrix) and C (measurement sensitivity matrix) of the statespace model are constructed. However, I have the following problem :
when designing the Kalman filter how should I compute the Q matrix (covariance matrix of the system noise), regarding the parameters given by matlab (e.g., K, strength of white noise)) ?
Thanks a lot for your help.
Catherine Marselli
Catherine Marselli Institute of Microtechnology Breguet 2 CH2000 Neuchatel
phone : + 41 32 718 34 10 fax : + 41 32 718 34 02 email : catherine.marselli@imt.unine.ch



