Abstract
Identification of systems realized with minimal order models is considered to be important. In this paper, a new identification method for linear discrete systems is presented.
Here the linear system is first divided into a subsystem and a noise process in order to identify the process dynamics and the noise characteristics separately. By applying a 3-level M-sequence to the subsystem and by observing its output, the Markov parameters of the subsystem are estimated, and a minimal order realization is accomplished. Consequently, the pulse transfer function of the subsystem can be estimated. The minimal order is found in the pulse transfer function of the noise process using Akaike's and Furuta's methods, where the autocorrelation functions of the noise process are introduced by the resultant pulse transfer function of the system.
Numerical examples confirm the adequacy of the proposed identification algorithm, and suggest that it is effective to use input signals having different SN ratios to achieve an accurate identification.
