A new iterative identification scheme for Hammerstein systems with support vector machine based on biconvex optimization

In this paper, we propose a new iterative algorithm using support vector machine (SVM) to identify Hammerstein systems based on biconvex optimization. The linear part of the system is allowed to be an infinite impulse response (IIR) system and the nonlinear static functions is a Borel measurable function including saturation nonlinearity, deadzone nonlinearity, quantization nonlinearity and signum nonlinearity. The algorithm is obtained by iteratively finding the minimum of a biconvex cost function. It is proved that under certain conditions the estimates generated from the algorithm converge to true parameters, which correspond to the global minimization of the cost function.