论文部分内容阅读
本文通过对频带受限数字信号的离散傅立叶变换特性的研究 ,引进了交替投影神经网络 ,并将其应用范围从实数域拓广到复数域 ,且给出了在复数域仍然成立的若干结论 .运用这些结论 ,在对网络噪声抑制、网络收敛速度及待外推信号因截断而造成频谱严重外泄等问题的分析与讨论的基础上 ,提出了一种基于交替投影神经网络的外推算法 .仿真实验表明该方法是行之有效的 .另外 ,该算法对频谱外推同样适用 ;由于它采用全互连神经网络结构 ,易于并行计算和VLSI实现 ,从而可满足军事上实时处理的需要 .
Based on the study of the discrete Fourier transform of band-limited digital signals, this paper introduces alternating projection neural network and expands its application range from real number field to complex number field, and gives some conclusions that still exist in the complex number domain. Based on the analysis and discussion of network noise suppression, network convergence speed and serious spectrum leakage due to truncation of the signal to be extrapolated, the paper proposes an extrapolation algorithm based on alternating projection neural network. Simulation results show that this method is effective. In addition, the algorithm is equally applicable to extrapolation of the spectrum. Because of its fully interconnected neural network structure, it is easy to be parallelized and implemented by VLSI to meet the needs of military real-time processing.