基于多主分量神经网络的同步DS-CDMA伪码盲估计

针对批处理方法在实现非等功率同步直接序列码分多址(direct sequence code-division multiple access,DS-CDMA)信号伪码序列盲估计时存在的复杂度高、收敛速度慢的问题,引入了3种多主分量神经网络(Sanger NN、LEAP NN和APEX NN)。首先将已分段的一周期DS-CDMA信号作为神经网络的输入信号,用神经网络各权值向量的符号函数代表DS-CDMA信号各用户的伪码序列,然后通过不断输入信号来反复训练权值向量直至收敛,最终DS-CDMA信号各用户的伪码序列就可以通过各权值向量的符号函数重建出来。此外,本文提出了一种在递归最小二乘(recursive least square,RLS)意义下的最优变步长收敛模型,极大地提高了网络的收敛速度。理论分析与仿真实验表明:将3种神经网络用于同步非等功率DS-CDMA信号伪码盲估计时的复杂度均明显降低,且LEAP NN与Sanger NN均可有效地实现-20dB信噪比、10个用户下的同步非等功率DS-CDMA伪码盲估计,APEX NN则相对较差,此外,LEAP NN消耗内存较大、收敛速度快,APEX NN相反,Sanger NN则介于两者之间。

基于Sanger神经网络的TDDM-BOC信号组合码序列盲估计

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。

Multivariate Regression Analysis on Correlated Characters about Fresh Pod Yield of Fresh Edible Soybean

Objective] The alm was to survey 10 characters of 8 fresh edibIe soy-bean varieties, analyze maln Ioading factors using principal component analysis, and estabIish muItipIe regression equation on fresh pod yield. [Methods] Through princi-pal component analysis on 10 characters of 8 fresh edibIe soybean varieties, char-acters reIated to fresh pod yield of fresh edibIe soybean were cIarified. [Results] Af-ter the principal components analysis, pod weight per pIant, 100-seed weight and pod number per pIant of fresh edibIe soybean were chosen to study their reIation with the yield of fresh edibIe soybean, moreover, it was demonstrated that the reIa-tion was Iinear reIation, thus it was suitabIe for muItivariate regression analysis. Fi-nal y, the mathematical expression formuIa about fresh pod yield was estabIished. [Conclusions] There were three characters affecting fresh pod yield, nameIy, pod weight per pIant, 100-seed weight and pod number per pIant, the mathematical equation was y=816.732＋4.145X6-0.718X8-0.985X9 （X6： pod weight per pIant; X8： 100-seed weight; X9： pod number per pIant）.