Skip to Main Content
In this paper, we introduce an adaptive algorithm for nonlinear system identification in the short-time Fourier transform (STFT) domain. The adaptive scheme consists of a parallel combination of a linear component, represented by crossband filters between subbands, and a quadratic component, which is modeled by multiplicative cross-terms. We adaptively update the model parameters using the least-mean-square (LMS) algorithm, and derive explicit expressions for the transient and steady-state mean-square error (MSE) in frequency bins for white Gaussian inputs. We show that estimation of the nonlinear component improves the MSE performance only when the power ratio of nonlinear to linear components is relatively high. Furthermore, as the number of crossband filters increases, a lower steady-state MSE may be obtained at the expense of slower convergence. Experimental results support the theoretical derivations.