Linear FIR precoding for frequency selective MIMO channels using Gaussian randomization

Finite impulse response (FIR) precoders are efficient precoding framework for frequency-selective multiple-input multiple-output (MIMO) channels. In this paper, the optimal linear FIR precoder is investigated to maximize the capacity of frequency-selective MIMO channel under average transmit power constraint. We use Kronecker product to regenerate the nonconvex optimization problem and propose a semidefinite programming (SDP) relaxation scheme with Gaussian randomization to provide approximate solutions. Numerical results indicate that the proposed algorithm outperforms the existing algorithms in the high signal to noise ratio (SNR) regime, especially when the FIR precoder has large filter taps.