In the long term evolution (LTE) uplink systems, reference signals based on Zadoff-Chu sequences are employed for time synchronization and block type channel estimation because the pilots are inserted into all the subcarriers of one single carrier-frequency division multiple access (SC-FDMA) symbol with a certain period. The block-type channel estimation can be adopted for slowly fading channel while the channel can be considered as stationary within a certain period of SC-FDMA symbols. However, in practice, a wideband radio channel is time-variant and the channel estimation errors based on block-type channel estimation will significantly degrade the performance of SC-FDMA systems in fast fading environments. Therefore, in this paper, the maximum likelihood (ML) time synchronization and comb-type channel estimation for SC-FDMA systems is presented. We demonstrated that the proposed system has significant bit error rate (BER) advantages over fast varying environments.