Abstract: This paper deals with a robust state estimation problem of descriptor systems subject to linear fractional disturbances, and proposes a Kalman type recursive filtering algorithm. It is proved that the linear fractional uncertainties can be equivalently represented by the intersection of a series of additive uncertainties. The problem of finding the least conservative set of additive uncertainties to approximate this intersection can be converted to a convex optimization one in the process of robust filtering. Numerical simulation results have also been reported which confirm the efficiency of the proposed algorithm. Moreover, for structured linear fractional uncertainties, the algorithm is shown to perform better than the available one.