遗传优化神经网络的水声信道盲均衡

不需要训练序列的盲均衡技术可以有效地节省水声通信带宽，消除码间干扰，提高水声通信效率和质量。以前馈神经网络（FNN）作为盲均衡器，既适用于最小相位信道，也适用于非最小相位信道，包括非线性信道，但是前馈神经网络在实际的应用中其网络拓扑结构的选取和初始权重的确定缺乏理论依据，且其训练主要依靠BP算法，存在收敛速度慢、容易陷入局部极值及“过学习”的问题。为此，本文提出了一种遗传优化神经网络的水声信道盲均衡算法（GA—BP），对前馈神经网络拓扑结构和网络权重同时优化，有效地克服了传统前馈神经网络盲均衡的缺陷，提高了前馈神经网络盲均衡的泛化性能并加强了跟踪时变信道的能力和对信道突变的适应能力。水池试验结果证明了文中提出的遗传优化神经网络水声信道盲均衡算法的有效性，与直接前馈神经网络盲均衡相比较，均衡性能明显得到了提高。

Blind equalization for underwater acoustic communication by genetic algorithm optimizing neural network

Blind equalization without training sequence is a bandwidth efficient way to mitigate the intersymbol interference in the field of underwater acoustic communication. Using FNN (Forward Feedback Neural Network) as blind equalizer can solve the equalization problems for minimum phase and non-minimum phase including non-linear channel. However, training of FNN is commonly based on BP algorithm, and thus suffers from slow convergence, local minimum point and excessive learning. In this paper, genetic algorithm optimizing neural network (GA-BP) is proposed as a new blind equalization method. Because of optimizing the topology of FNN and the weights of network simultaneously, the new algorithm overcomes the default of traditional FNN blind equalization and enhances the ability of tracing the time-varying channel and adapts to channel mutation. Results from experimentation in a channel pool indicate that using the algorithm proposed in this paper provides effectiveness. Compared with FNN, the performance of equalization is improved obviously.