Blind identification of multichannel systems driven by impulsive signals

We study the problem of estimating the impulse responses of multiple FIR channels driven by an unknown impulsive signal. A deterministic method which deals with arbitrarily short excitation signals is proposed by utilizing subspace properties of the data covariance matrix. A variant of the proposed method is also presented which exploits the L1-norm convex optimization to improve the estimation accuracy when the impulse responses of the system to be identified are sparse. The proposed methods have the potential to improve the performances of communication, exploration seismology and mechanical signature analysis systems.