An ICA-based adaptive filter algorithm for system identification using a state space approach

This paper proposes a new ICA-based adaptive filter algorithm for system identification using a state space approach. An additive noise model is considered and the signal is separated from the noisy observation. First, we introduce an augmented state-space expression of the observed signal representing the problem in terms of ICA, and then using the natural gradient, we derive a new algorithm. The local convergence conditions of the proposed algorithm is derived. Some simulations are carried out to illustrate its effectiveness.