Phase-shifted interpolation for channel matrix inversion in MIMO-OFDM systems

Channel matrix inversion, which requires significant hardware resource and computational power, is a very challenging problem in MIMO-OFDM systems. Casting the frequency-domain channel matrix into a polynomial matrix, interpolation-based matrix inversion provides a promising solution to this problem. In this paper, by showing that the polynomial coefficients can be well approximated by a Gaussian function, we propose an efficient algorithm, which relaxes the requirement for knowing the maximum multipath delay spread and enables the use of simple low-complexity interpolators by introducing a phase shift term to the signal to be interpolated. Simulation results show that significant complexity saving can be achieved with little equalization performance degradation.