本文在分析了当前主要混沌神经元网络模型及其在信息处理中的应用研究的基础上，进行相应的改进，提出了一维双向耦合神经元网络和一种改进的自联想矩阵的定义方法。一维双向耦合神经元网络克服了全局耦合网络的缺点但同样具有全局耦合映象的分类功能并可以推广到许多信息处理应用中。通过仿真实验证实了，采用改进的自联想矩阵定义方法建立的混沌神经元联想记忆网络。与现有方法相比，其联想的成功率和回忆速度得到了较大的提高。并从理论上详细地分析了网络的联想记忆过程和性能提高的原因。 In this paper, based on the analysis of the current main chaotic neural network model and its application in information processing, a corresponding improvement is made. One-dimensional bidirectional coupled neural network and an improved definition method of self-associative matrix are proposed. One-dimensional bidirectional coupled neuron networks overcome the shortcomings of the global coupling network but also have the classification function of the global coupling map and can be extended to many information processing applications. Through the simulation experiments, we prove that the associative memory network of chaotic neurons established by the improved self-associative matrix definition method. Compared with the existing methods, the association’s success rate and memory speed has been greatly improved. And theoretically detailed analysis of the network associative memory process and performance reasons.