A recursive algorithm for the two-stage empirical frequency-domain optimal param- eter(EFOP)estimation method was proposed.The EFOP method was a novel sys- tem identification method for Black-box models that combines time-domain esti- mation and frequency-domain estimation.It has improved anti-disturbance perfor- mance,and could precisely identify models with fewer sample numbers.The two- stage EFOP method based on the boot-strap technique was generally suitable for Black-box models,but it was an iterative method and takes too much computation work so that it did not work well online.A recursive algorithm was proposed for dis- turbed stochastic systems.Some simulation examples are included to demonstrate the validity of the new method. A recursive algorithm for the two-stage empirical frequency-domain optimal param-eter (EFOP) estimation method was proposed. The EFOP method was a novel sys- tem identification method for Black-box models that combines time-domain estimation and frequency -domain estimation. It has improved anti-disturbance perfor- mance, and could precisely identify models with fewer sample numbers. The two-stage EFOP method based on the boot-strap technique was generally suitable for Black-box models, but it was an iterative method and takes too much computation work so that it did not work well online. A recursive algorithm was proposed for dis- turbed stochastic systems. Some simulation examples are included to demonstrate the validity of the new method.