Efficient interpolation of precoding matrices in MIMO-OFDM systems

Abstract

One way of fully exploiting the channel capacity in multiple-input multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) systems is to deploy precoding vectors derived from SVD to form parallel data pipes. Optimal precoding requires accurate channel state information for each OFDM subcarrier. This paper proposes a limited channel decomposition architecture by using an interpolation method to reduce the computation cost of precoding. In this scheme, only on the pilot subcarriers the estimated channel information is used to get SVD decomposition results, and precoding vectors on other subcarriers are obtained by an orthogonal interpolation method. The interpolation relies on an important assumption that the singular modes on adjacent subcarriers change approximately in a linear way. Numerical simulations suggests that with appropriate settings, e.g., proper FFT size and interpolation gap etc., the proposed scheme can reduce computation efficiently while only sacrifice little performance.