摘要： 利用遗传算法全局搜索能力强和反向传播(BP)算法局部搜索速度快的特点，采取两段式训练方法，既避免陷入局部最小，又加快收敛速度。提出基于遗传算法优化神经网络权值的多用户检测算法。采用实数编码方式，将传统神经网络的能量函数作为适应度函数，选择算子选用轮盘赌算子，交叉算子选用单点交叉算子，变异算子选用正态变异算子。仿真结果表明，该算法的误码率、信干比和信道跟踪能力等方面的性能与传统前馈神经网络多用户检测算法相比均有一定的改善。

关键词: 多用户检测, 遗传算法, 神经网络, 收敛速度, 误码率

Abstract: Using genetic algorithm global search capability and Back Propagation(BP) algorithm for local search and fast speed, two-stage training methods are taken. The two both not only avoid falling into local minimum, but also speed up the convergence speed. Multi-user detection algorithm based on optimization neural network weight with GA is proposed. In the algorithm, use real encoding, choose the traditional neural network energy function as fitness function, selection operator selection of roulette operator, crossover operator selection of one-point crossover operator, mutation operator chosen normal mutation operator. Simulation experiments show that the algorithm’s bit error rate, signal to interference ratio, channel tracking capability and performance have significant improvement compared with traditional feed forward neural networks.

Key words: multi-user detection, genetic algorithm, neural network, convergence rate, bit error rate

中图分类号: 

• TP301.6