This paper proposes an algorithm to estimate external forces exerted on the end-effector of a robot manipulator using information from joint torque sensors (JTS). The algorithm is combination of Time Delay Estimation (TDE) and input estimation technique where the external force is considered as an unknown input to the robot manipulator. Based on TDE's idea, the estimator which does not require an accurate dynamics model of the robot manipulator is developed. The simultaneous input and state estimation is used to reject not only nonlinear uncertainties of the robot dynamics but also the noise of measurements. The performance of the proposed estimation algorithm is evaluated through simulation of a two degree-of-freedom manipulator and it demonstrates the stability in estimating the external forces. The estimation results show that this approach allows inexpensive sensors as joint torque sensors to be used instead of expensive ones as F/T sensors in robot application.