Independent component analysis for nonminimum phase systems using H∞ filters

This paper proposes an independent component analysis (ICA) method using H∞ filters for nonminimum phase systems. In the basic ICA approach, the input signal is recovered by estimating the parameter of the inverse of the mixing system. If the system is nonminimum phase, the estimated parameter diverges due to the instability of the inverse. For this problem, a inverse filter is constructed based on an H∞ filter in order to estimate the state of the given plant. The learning algorithm to estimate the parameter of the system is derived by minimizing the Kullback-Leibler divergence. Furthermore, a numerical simulation demonstrates the effectiveness of the proposed method.