基于Sanger神经网络的TDDM-BOC信号组合码序列盲估计

发布时间：2018-06-25 08:30

本文选题：神经网络 + 多主分量　；参考：《计算机应用》2017年08期

【摘要】：针对低信噪比(SNR)下时分数据调制二进制偏移载波调制信号(TDDM-BOC)的组合码序列盲估计问题,提出一种基于Sanger神经网络(Sanger NN)的新方法。首先将已分段的信号作为输入信号并利用Sanger NN提取各主分量的权值向量;然后通过其多次输入反复训练权值向量,直至权值向量达到收敛;最终利用各个权值向量的符号函数重建信号的组合码序列,实现TDDM-BOC组合码序列的盲估计。此外,采用最优变步长的方法来提高收敛速度。理论分析和仿真实验表明,Sanger NN可以实现-20.9~0 d B信噪比下TDDM-BOC信号组合码序列的盲估计,且其复杂度明显低于传统奇异值分解(SVD)法和自适应特征提取的在线无监督学习神经网络(LEAP);尽管Sanger NN收敛所需数据组数大于LEAP,但收敛时间明显少于LEAP算法。
[Abstract]:A new method based on Sanger neural network (Sanger NN) is proposed for blind estimation of binary offset carrier modulated signals (TDDM-BOC) based on time-division data modulation (TDDM-BOC) under low signal-to-noise ratio (SNR). Firstly, the segmented signal is taken as the input signal and the weight vector of each principal component is extracted by using Sanger NN, then the weight vector is trained repeatedly through its multiple inputs until the weight vector converges. Finally, the signal combination code sequence is reconstructed by the symbol function of each weight vector, and the blind estimation of TDDM-BOC combination code sequence is realized. In addition, the optimal variable step size method is used to improve the convergence rate. Theoretical analysis and simulation experiments show that Sanger NN can realize blind estimation of TDDM-BOC signal combination code sequences under -20.9 ~ 0 dB SNR. Its complexity is obviously lower than that of the traditional singular value decomposition (SVD) method and the on-line unsupervised learning neural network (LEAP) based on adaptive feature extraction, although the number of data sets required for Sanger NN convergence is larger than that of LEAPs, the convergence time is obviously less than that of LEAP algorithm.
【作者单位】：重庆邮电大学信号与信息处理重庆市重点实验室;
【基金】：国家自然科学基金资助项目(61671095,61371164) 信号与信息处理重庆市市级重点实验室建设项目(CSTC2009CA2003) 重庆市教育委员会科研项目(KJ130524,KJ1600427,KJ1600429)~~
【分类号】：TN911.23;TP183

【参考文献】