Time-Varying Channel Estimation for MIMO/OFDM Systems Using Superimposed Training

This paper deals with the problem of time-varying (TV) channel estimation for multiple-input multiple-output/orthogonal frequency-division multiplexing (MIMO/OFDM) systems based on superimposed training (ST). The time-varying coefficients of the TV channel are firstly modeled by truncated discrete Fourier bases, and then optimally estimated both in one OFDM symbols and over multiple OFDM symbols by judiciously designing the superimposed pilots. In addition, an iterative symbol detection based channel estimation scheme with analytical performance analysis, is provided to mitigate the interference due to the unknown information sequences (thus to further improve channel estimation performance as well as SER levels). Simulations confirm that the proposed estimator achieves a considerable gain in estimating TV channels, and exhibits a nearly indistinguishable symbol error rate performance from the OFDM systems of frequency-division multiplexed trainings.