Blind SIMO-FIR second order identification: a robust approach

This paper addresses the problem of the second order blind identification of single input multiple output (SIMO) finite impulse response (FIR) systems. The advantage of using the second order methods lies in rather weak assumptions on the system structure and the properties of the input signal. Only a good diversity of measurements at different outputs is required to achieve high estimation accuracy when the sample size is limited. However an important drawback of such methods is the high sensitivity to the order modeling errors. The known methods which are robust in the presence of order modeling errors exploit a supplementary information on the data model. A new approach to the adaptive order selection, proposed in this paper requires no supplementary knowledge about the systems nor the input signals. The author presents a particular deterministic version of this general approach which yields perfect identification in the noiseless case.