A cumulant subspace approach to FIR multiuser channel estimation

Blind identification of FIR multiuser channels using higher-order statistics (HOS) is investigated in this work. Higher order statistical information is exploited to identify systems with an equal number of users and outputs. A sufficiency condition is presented and is less stringent than many known conditions. Furthermore, a new blind identification method is developed which extends the second order subspace approach to HOS with closed-form solutions. This algorithm is shown to be capable of identifying a wider class of channels and is robust to some channel order over-estimation