一种递归神经网络空间分集均衡器

水声信道中的衰变多途特性常引起难以克服的码间干扰(ISI)。为了降低误码率、改善系统性能,本文提出了一种递归神经网络空间分集均衡器。它在传统的空间分集结构中融入了多层感知器,该结构能够充分利用训练信号的信息,在可调权数较少的情况下,能够得到较好的均衡效果;针对该结构,采取了初始设定权值的办法,从而使该均衡器的收敛速度得到大大提高。仿真与湖试试验结果表明,该均衡器结构合理,对空间分集均衡性能有一定的改善。

A recurrent neural network spatial diversity equalizer

The fading and multipath underwater acoustic channels have always been great impediments to building reliable underwater communication systems. In order to improve the system performance and decrease the error probability a structure called the Recurrent Neural Network Spatial Diversity Equalizer (RNNSDE) is studied in this paper. We put a two-layer neural network into the conventional structure of the spatial diversity equalizer. The proposed structure can take full advantage of the training information, and better performance can be gained with a small amount of weights. By setting initial values of the adaptive weights for the structure, rapid convergence can be obtained. The simulation and field test results show that the proposed structure is reasonable and it can improve the performance of the spatial diversity equalizer.