基于随机Hopfield神经网络的最优多用户检测器

Optimal Multiuser Detectors Based on the Stochastic Hopfield Network

Hopfield神经网络检测器与传统检测器相比具有独特的魅力 ,但存在局部收敛问题 .随机Hopfield神经网络通过对状态变量引入噪声 ,从而避免局部收敛 ,进而达到最优检测 .本文给出了CDMA移动通信系统中基于随机Hopfield神经网络的最优多用户检测器的统一框架 ,该检测器利用随机Hopfield神经网络对最大似然函数进行优化从而得到最优检测性能 .在三种典型信道中对这些检测器性能进行比较 ,结果表明该检测器比传统检测器的性能更好 .

The detector using Hopfield network is notable compared with traditional detectors, but it has local optimal behavior. The stochastic Hopfield network is the network, where noise is added to the state variables in order to reach the global optimum and achieve the best detection. The uniform theoretic framework of the detectors based on the stochastic Hopfield network in CDMA communication system is presented. The novel detector guarantees optimal performance by minimizing the corresponding quadratic form. The performance of the novel detector has been tested by extensive simulations in three typical channels, which justifies its superiority compared with other traditional methods.