基于神经网络的变换域通信系统接收技术

Neural network based receiving technology in transform domain communication system

提出了一种不须要估计本地基函数的变换域通信系统接收机接收方法.详细分析了接收机的接收原理,以相关解调信号能量最大化为目标构建了接收机最优化接收模型,通过对模型求解得到了接收数据自协方差矩阵的最大特征值向量,此向量可作为接收端基函数实现最佳接收,从而引入基于Hebbian学习训练的神经网络完成最大特征值向量的自适应估计与数据接收.仿真结果表明:当算法收敛时,作为基函数估计的网络连接强度矢量与发送端基函数基本一致,并且当发送端改变基函数时,算法依然能够实现准确跟踪.系统的接收性能与收发两端基函数一致条件下的接收性能基本一致,当算法迭代次数为300,误码率为1×10-4时,系统的信噪比损失可小于1dB.

A technique of the receiver without estimation of the local basis function in transform do-main communication system （TDCS） was proposed .A detailed description of the principle of the re-ceiver was analyzed .The objective of the newly established model of receiver is to maximize the ener-gy of the correlation demodulating signal ,and the maximum eigenvector solved from the covariance matrix of the reception data can be regarded as the basis function of the optimal receiver .T he neural network（NN） based on Hebbian learning and training process was introduced to adaptively estimate the maximum eigenvector .Theoretical analysis and simulation results show that the network connect strength vector of the estimated basis function is quite similar to that of the transmitter when the algo-rithm is convergence , and even w hen the basis function of transmitter varies , the algorithm can achieve accurate tracking as well .The reception performance of the system is in good agreement with that under the ideal condition when the number of iterative detection is 300 with a bit error rate of 1 ＆#215; 10-4 ,and the signal-to-noise ratio （SNR） loss is below 1 dB .