The objective of our study is to build a precise model by applying the technique of system identification for the model-based control of a nonlinear robot arm, taking joint-elasticity into consideration. This paper proposes a systematic identification method, called "decoupling identification", for a serial two-link robot arm with elastic joints caused by the Harmonic drive ® reduction gears. The proposed method serves as an extension of the conventional rigid-jointmodel-based identification. The robot arm is treated as a serial two-link two-inertia system with nonlinearity. The decoupling identification method using link-accelerometer signals enables the serial two-link two-inertia system to be divided into two linear one-link two-inertia systems. The MATLAB®'s commands for state-space model estimation are utilized in the proposed method. Physical parameters such as motor inertias, link inertias, joint-friction coefficients and joint-spring coefficients are estimated through the identified one-link two-inertia systems. Experimental results using a SCARA-type planar two-link robot arm with elastic reduction gears showed an accuracy of the proposed identification method.