A new direct approach to identifying the parameters of distributed paramete
r systems from noise-corrupted data is introduced. The model of the system
which takes the form of a set of linear or nonlinear partial differential e
quations is assumed known with the exception of a set of constant parameter
s. Using finite-difference approximations of the spatial derivatives the or
iginal equation is transformed into a set of ordinary differential equation
s. The identification approach involves smoothing the measured data and est
imating the temporal derivatives using a fixed interval smoother. A least-s
quares method is then employed to estimate the unknown parameters. Three ex
amples that illustrate the applicability of the proposed approach are prese
nted and discussed.