This paper proposes a composite control approach based on the robust feedback linearization and Extended Kalman Filter (EKF) to improve the tracking performance for the flexible link manipulators in the presence of torque disturbances. In this regard, first an EKF is employed to estimate the disturbances, utilizing the uncertain nonlinear model of flexible link system in addition to noisy measurement data. The estimate is then used in the control strategy in order to eliminate the destructive effects of the disturbances. It can also be used as interaction data in some applications such as telerobotics. In the next step, a Lyapunov Redesign Feedback Linearization (LRFL) approach is utilized in order to alleviate the effect of model uncertainties, disturbance estimation error as well as nonlinearities presented in the dynamics of the redefined output. This output is selected as a point close to the tip to avoid the difficulties associated with the non-minimum phase behavior of the tip position. Simulation results performed on a single-link flexible manipulator are presented to illustrate the significant capability of this technique in tracking performance as well as disturbance estimation ability in uncertain and noisy situations.