Two-dimensional, frequency domain, adaptive system modeling using 3-D spatiotemporal inputs

An adaptive, two-dimensional (2-D) frequency domain system modeling technique is presented. This technique is implemented in the 3-D spatiotemporal domain. The algorithm derived is based on the recursive like structure, and models the unknown 2-D spatially linear and invariant system by a 2-D autoregressive moving-average (ARMA) structure. For each pair of input and output images of the 2-D unknown system, the algorithm, which uses the steepest descent method, minimizes the error function in the least-square sense by adjusting the coefficients of the 2-D ARMA model at each iteration. The homogeneous convergence factors associated with this algorithm are developed. Computer simulations demonstrate that this algorithm, with the derived homogeneous convergence factor, has excellent adaptation accuracy and convergence speed. This algorithm is successfully applied to modeling a time-varying 2-D system