In this paper, a new H-infinity filter design approach is proposed for discrete-time linear systems with unknown inputs. The proposed Lyapunov-based approach, free of any similarity transformations, designs a linear filter for the modified model of the plant, and then extracts the original states of the system. The designed filter estimates both the system states and the unknown input simultaneously and does not have many of the restrictive assumptions and restrictions that the existing unknown input filters do. In the end, simulation results are used to illustrate the effectiveness of the proposed filter. (C) 2014 Elsevier Ltd. All rights reserved.