实现最小互熵编码多用户检测的快速算法

Fast algorithm to implement minimum cross entropy coded multiuser detection

基于最小互熵(minimumcrossentropy,MCE)的迭代多用户检测算法在高度互相关的系统(即非扩频系统)中可以渐近地获得单用户的性能,是一种最优检测算法。但标准的MCE算法的计算复杂度极高,因而使其难以实用。该文提出了一种实现MCE算法的快速算法,避免了标准MCE算法在计算度量函数时的重复运算,在用户间符号同步和异步两种情况下均可以将计算复杂度降低K倍(K为用户数),而无性能损失。对于异步系统,还可以此基础上,通过对度量函数作合理近似,使运算度进一步减少1/3,在加性高斯白噪声信道下的仿真结果表明,这种近似所产生的性能损失低于0.2dB。

The performance of an algori th m derived from iterative techniques for minimizing the cross-entropy (MCE) can asymptotically approach single-user performance even in highly correlated multi user systems when the MCE algorithm optimally detects multiuser signals. However , direct implementation of the MCE algorithm is too complex. Therefore, a fast a lgorithm was developed to implement the MCE algorithm. The algorithm avoids redu ndancy when calculating the metric function, so it can reduce the complexity by K times (K is the number of the users) to keep the performance from deg rading for both synchronous and asynchronous operation. For asynchronous system, the algorithm can further reduce the calculations by about 1/3 by modifing the MCE algorithm metric function. Simulations show that the loss in performance by approximating metric is less than 0.2 dB compared to the MCE detector for a channel with additive white Gaussian noise.