This paper studies the parameter identification of stochastic systems with colored noises. Using the data filtering technology to filter the input and output data, which converts the original system with moving average noise to the system with white noise, we propose the filtering-based extended stochastic gradient algorithm and analyze its convergence. In addition, in order to improve the parameter estimation accuracy and accelerate the convergence rate, a filtering-based multi-innovation extended stochastic gradient algorithm is proposed by using the multi-innovation identification theory and its convergence is analyzed. Compared with the extended stochastic gradient algorithm, the proposed filtering-based extended stochastic gradient algorithm and the filtering-based multi-innovation extended stochastic gradient algorithm can obtain higher precision parameter estimates. Finally, the simulation results indicate that the proposed algorithms are effective.