文中研究了广泛应用于多输入多输出雷达系统中的正交相位编码集的设计问题,提出了一种基于复值神经网络的正交相位编码集设计方法。该方法将编码集设计视为非线性系统构建问题,通过最小化反映了编码集自相关和互相关性质的损失函数,在常模约束条件下对网络误差进行后向传播,从而通过网络训练获得具有良好相关性的正交相位编码集。通过与现有先进算法进行不同条件下的仿真对比,文中方法设计得到的正交相位编码集具有更好的相关特性,充分验证了本文所提方法的先进性。

In this paper, the task of designing constant modulus orthogonal phase code sequence set is considered, which is widely applied in multiple-input multiple-output radar systems. By regarding the sequence sets as non-linear systems, a novel method based on complex-valued neural network is proposed to deal with the design task. The proposed method minimizes the loss function which reflects the auto-correlation and cross-correlation functions of sequence sets and carries out back propagation process with constant modulus constraint, so that orthogonal phase code sequence sets can be obtained through network training. Compared with the existing advanced algorithms under different conditions, the orthogonal phase code set designed by the proposed method has better correlation characteristics, which fully verifies the progressiveness of the proposed method.

TN957.51