复数Hopfield盲恢复多用户QPSK信号

为盲恢复MIMO系统QPSK信号,构造了一个专门的复数Hopfield神经网络(CHNN).证明了CHNN趋向平衡的充要条件,指出了虚、实部同时平衡是极值点的必要条件,给出了平衡点集大小与连接矩阵主对角元大小之间的关联.此外,为准确地串行恢复多用户信号,提出了新的补空间扩展法.计算机仿真表明,该CHNN所采用的连接矩阵和补子空间扩展法能有效地削减平衡点数,减少神经网计算次数,降低用户信号漏检的错误.

Blind recovery of multiuser QPSK with complex Hopfield network

A complex Hopfield neural network(CHNN) with a special connective matrix is constructed to blindly recover QPSK(quadriphase-shift keying) signals of MIMO(multiple-input multiple-output) systems.A necessary and sufficient condition of the CHNN equilibrium is also proposed,which indicates that an equilibrium point is a likely minimal point only if both of the real part and imaginary part of CHNN states reach equilibrium synchronously,and it is shown that diagonal elements of the CHNN connective matrix have an effect on the number of equilibrium points.In addition,a new method of a complementary subspace extension is presented to accurately recover multiuser signals in serial.Computer simulations show that the special connective matrix is helpful to reduce the number of equilibrium points and computation cost,and the new method of a complementary subspace extension is beneficial to decrease detection miss of multiuser signals.