Several diversity techniques have been proposed to cope with fading in wireless networks. A recent and promising technique, which does not increase hardware demands, is cooperative communication. Here spatial diversity is created by using other terminals to relay the information from the source terminal to the destination terminal. Cooperative protocols proposed in literature are often studied under the simplifying assumption that all channel state information is available at the destination. This contrasts to a practical scenario where channel estimates need to be derived from the broadcasted signals. In this paper, we focus on the Amplify-and-Forward protocol and use the Space-Alternating Generalized Expectation-maximization (SAGE) algorithm to estimate the channel gain and noise variance in an iterative way, exploiting both the pilot part and the data part of the signal received at the destination. We consider a relay that performs a general affine transformation on the pilot symbols received by the source, and we optimize this transformation in terms of estimation accuracy. The resulting mean square error performance is shown to be close to the Cramer-Rao lower bound (CRB).