ADAPTIVE, FREQUENCY-DOMAIN, 2-D MODELING USING SPATIOTEMPORAL SIGNALS

In this paper, an adaptive, frequency domain, steepest descent algorit hm for two-dimensional (2-D) system modeling is presented. Based on th e equation error model, the algorithm, which characterizes the 2-D spa tially linear and invariant unknown system by a 2-D auto-regressive, m oving-average (ARMA) process, is derived and implemented in the 3-D sp atiotemporal domain. At each iteration, corresponding to a given pair of input and output 2-D signals, the algorithm is formulated to minimi ze the error-function's energy in the frequency domain by adjusting th e 2-D ARMA model parameters. A signal dependent, optimal convergence f actor, referred to as the homogeneous convergence factor, is developed . It is the same for all the coefficients but is updated once per iter ation. The resulting algorithm is called the Two-Dimensional, Frequenc y Domain, with Homogeneous mu, Adaptive Algorithm (ZD-FD-HAA). In add ition, the algorithm is implemented using the 2-D Fast Fourier Transfo rm (FFT) to enhance the computational efficiency. Computer simulations demonstrate the algorithm's excellent adaptation accuracy and converg ence speed. For illustration, the proposed algorithm is successfully a pplied to modeling a time varying 2-D system.