文章在一种基于多维离散傅立叶变换原理的前馈神经网络模型的基础上，提出了一个适用于高精度逼近和泛化建模的神经网络。它可用来逼近任何连续函数，且逼近精度高，泛化能力强，学习速度快。计算机模拟实验结果显示了该网络较好地利用了离散傅立叶交换在结构和性能上的优点，在天线罩视线误差校正建模研究上很好的效果。 Based on a feedforward neural network model based on the principle of multidimensional discrete Fourier transform, a neural network suitable for high precision approximation and generalization modeling is proposed. It can be used to approximate any continuous function, and the approximation accuracy, generalization ability, learning speed. The results of computer simulation show that this network can make good use of the advantages of discrete Fourier transform in structure and performance, and it has a good effect in the calibration research of antenna line of sight error correction.