This paper describes a new robust fault detection strategy to detect stator winding fault in induction motors. Because of the destructive and rapidly propagation of stator winding fault, it is critical to detect it as soon as possible. The proposed approach is based on Unknown Input Observer (UIO) and Extended Kalman Filter (EKF). The EKF and UIO have been used for speed estimation and fault detection purpose respectively. A nonlinear UIO that can detect fault in induction motors is proposed in this paper. In the UIO design procedure, for flexible design, the stator currents and speed of IM are assumed as measurement signals. After designing nonlinear UIO, an EKF is used for speed estimation of IM. Hence, the speed measurement is released. Finally, simulation results are shown for winding stator fault with various inter-turn faults and the validity of proposed approach has been verified.