Skip to Main Content
Assuming perfect channel state information, the existing interference alignment (IA) algorithm proposed in  suppresses inter-cell interference (ICI) by aligning ICI to a randomly selected reference vector. However, IA in practice relies on limited feedback, resulting in residual ICI. In this letter, we propose the optimization of the reference vector for regulating the residual ICI. Specifically, it is shown that the reference vector that minimizes an upper bound on the residual ICI power is the eigenvector corresponding to the largest eigenvalue of the sum of the interference-channel matrices multiplied by their corresponding Hermitian matrices. Moreover, the performance gain of the proposed IA algorithm compared with the existing one in  is analyzed and demonstrated by simulation to be significant.