Abstract: The channel state information (CSI) of all links should be available at all nodes for cooperative strategies design in relay networks. However, many channel estimation methods for amplify-and-forward (AF) relay networks only estimate a compound CSI of source-relay-destination (S-R-D) link. In this paper, we propose a channel estimation scheme for destination to disintegrate the CSIs of S-R and R-D links from the cascaded S-R-D link in AF relay networks. The proposed scheme estimates the CSI of R-D link firstly by utilizing some pilot signals inserted at the relay node. Then, the CSI of S-R link is estimated based on the received pilot signals and the pre-estimated CSI of R-D link. The channel estimation error of the proposed scheme is also analyzed. Especially, least square (LS) algorithm and minimum mean square error (MMSE) algorithm for the CSI estimation of S-R link are compared. We verify that the mean square error of LS estimation for S-R link is nonconvergent but that of MMSE estimation is convergent. At the end of this paper, simulation results are given to support the theoretical results.