一种有限字符输入的线性预编码设计算法

摘要

Addressing the problem of maximizing the mutual information of MIMO channels with finite alphabet inputs, a low-complexity algorithm of linear precoding is designed. According to the mercury/water-filling theory, the algorithm integrates the precoding method based on the uniformly rotated Space-Time Linear Constellation Precoding (ST-LCP) matrix and the method of maximizing the minimum distance between output vector signals, where the one with higher mutual information is chosen for precoding. Then, in the precoding technique based on the uniformly rotated ST-LCP matrix, the singular value matrix of MIMO channels is selected to be the power allocation matrix, and two modifications are designed including the local search and the power addition on the singular value matrix. Finally, the computational cost of mutual information is further decreased by exploiting the symmetry of the finite alphabet set. The proposed algorithm obtains mutual information close to the theoretical maximum under various channel and SNR conditions, and reduces or even avoids search, resulting in much lower computational complexity. The simulation results verify the effectiveness of the proposed algorithm.%针对有限字符输入下多输入多输出(MIMO)信道的互信息最大化问题，该文提出一种复杂度低的线性预编码算法。该算法根据水银/注水理论，融合基于均匀旋转的空时-线性星座预编码(Space-Time Linear Constellation Precoding, ST-LCP)矩阵的预编码方法和最大化最小输出向量信号间距的方法，从两者中选择互信息更高者用于预编码。然后，在基于均匀旋转的ST-LCP矩阵的预编码方法中，把MIMO信道的奇异值矩阵作为功率分配矩阵，并提出局部搜索和矩阵加幂两种改进措施。最后，利用有限字符集的对称性，进一步降低了互信息的计算复杂度。该算法在各种信道和信噪比条件下均能逼近互信息的理论最大值，并且减少甚至避免了搜索，计算复杂度大大下降。仿真结果验证了该算法的有效性。