In this study, the authors present a novel channel estimation scheme for a two-hop non-regenerative multiple-input multiple-output relay system. The proposed scheme has two stages. In the first stage, the source node transmits channel training sequence to the relay node. The relay node amplifies the received signal and forwards it to the destination node. During the second stage, the relay node amplifies the pre-received signal with another amplification matrix, and forwards it to the destination node. Based on the proposed scheme, by using parallel factor (PARAFAC) analysis, they further propose a low complexity channel estimation algorithm, where full knowledge of all channel matrices involved in the communication can be estimated at the destination node. Compared with existing algorithms, the proposed algorithm yields smaller channel estimation error with lower complexity. Numerical examples are shown to demonstrate the effectiveness of the low complexity PARAFAC-based channel estimation algorithm.