Abstract: Due to the difficulty in obtaining real-time statistics of time-varying channels, a multi-symbol channel estimation and data detection algorithm based on the expectation-maximization (EM) iteration was proposed for OFDM systems. In order to obtain a proper initial estimate for the iteration, a LS-based low-complexity scheme was implemented by making use of comb-type pilots. With the relationship between the channel statistics and signal to noise ratio in the proposed algorithm, a simplified algorithm with the channel autocorrelation matrix ignored was also introduced, and hence the operation of estimating channel statistics was avoided. Simulation results validate its robust properties to the absence of exact channel statistics such as power delay profile and Doppler power spectrum, and show that the normalized mean squared error and the bit error rate can be significantly reduced by this algorithm without exact Doppler power spectrum.