Evaluation of the Mutual Information of MIMO Channel with Finite-Alphabet Inputs

Mutual information (MI) between channel inputs and outputs plays a critical role in communication systems. However, with finite-alphabet inputs, no closed-form solution exists to the MI of multiple-input multiple-output (MIMO) channel. Although numerical integration methods can be used to estimate MI, it can be quite slow, especially for larger input-output dimensions. This paper presents a new way to approximately evaluate the MI of MIMO channel with equiprobable M-ary inputs. With the aid of entropy power Gaussian approximation, a MIMO channel can be approximately decomposed into parallel independent single-input multiple-output (SIMO) channels which are, in terms of MI, equivalent to single-input single-output (SISO) channels. Thus the MI of MIMO channel can be approximately evaluated with that of multiple SISO channels. With this result, the computational complexity can be reduced to linear complexity. Furthermore, we proof that the proposed approximation gives the true MI with massive receiver antennas. Simulation results verify the accuracy and effectiveness of the proposed method.