Fractionally differenced Gaussian noise（fdGn） process is a discrete time equivalent of fractional Brownian motion.Filtered versions of such processes are ideally suited for modeling signals with different short-term and long-term correlation characteristics.In this paper

we study the structure identification of filtered fdGn processes.For signals that exhibit a weak or moderate long-term persistence

an OIVPM-MDL structure identification method is proposed based on overdetermined instrumental variable product moment（OIVPM） and minimum description length（MDL） criteria

and for signals with a strong long-term correlation between samples a combinative approach of inverse fractional filtering and the OIVPM-MDL method is also described.Examples are given to illustrate the feasibility and effectiveness of the proposed method.