Neumann Series Expansion Based LMMSE Channel Estimation for OFDM Systems

In an orthogonal frequency division multiplexing (OFDM) system, the linear-minimum-mean-square-error (LMMSE)-based channel estimator often requires a matrix inversion with cubic complexity. In this letter, by employing a K terms Neumann series expansion to approximate the matrix inversion, the computational complexity is reduced to O(N log L) per channel realization, where N is the number of subcarriers and L is the number of non-zero time domain channel taps. Extensive simulation results show that even with small K (K ≤ 2), the performance loss caused by the proposed approximation is marginal.