Doubly dispersive channel estimation with scalable complexity

In this paper, we present an Approximate Linear Minimum Mean Square Error (ALMMSE) fast fading channel estimator for Orthogonal Frequency Division Multiplexing (OFDM). The ALMMSE channel estimator utilizes the knowledge of the structure of the autocorrelation matrix given by the Kronecker product between the time correlation matrix and the frequency correlation matrix. We separate the Linear Minimum Mean Square Error (LMMSE) filtering matrix into two matrices corresponding to individual filtering in frequency and time. The eigenvalues of these two matrices are rank-one approximated by the eigenvalues of the LMMSE filtering matrix. The complexity of the ALMMSE estimator can be scaled by varying the number of the considered number of eigenvalues. Simulation results show that the proposed ALMMSE channel estimator looses only 0.1 dB compared to the LMMSE channel estimator in realistic scenarios.