| System identification enables the construction of mathematical models using a system's observed input and output data. This thesis uses this approach to create mathematical representations of systems that are both nonlinear and time-varying. A nonlinear system model, the Hammerstein cascade comprising a static nonlinearity followed by a linear filter, has been extensively studied in this thesis. This system model has been used efficiently to model systems in control, communication and bio-medical applications. The identification approach in this thesis was developed with conventional techniques in mind, thus the system's time variations were transformed using sets of temporal basis sequences. This results in a time-invariant identification problem with respect to the expansion parameters. A parsimonious model of the system was obtained by selecting significant parameters using a bootstrap term selection technique. Applications of the developed algorithm to both simulated and experimental data were used to demonstrate its performance. |
