中文摘要：

为了探索无线信道的传播特征并建立相应的无线信道模型,针对无线信道的“指纹”特征建模及场景识别问题,提出基于无迹卡尔曼神经网络的信道建模方法.首先采用霍特林变换对信道传播数据进行预处理,实现数据从复数域到实数域的转换,然后利用主元分析对数据进行降维,得到降维数据后,采用无迹卡尔曼神经网络进行无线信道“指纹”特征建模,最后根据模型的输出值建立无线信道的评价指标,对测量的信道数据进行场景划分以及场景识别.仿真结果表明：无迹卡尔曼神经网络所建信道模型可准确对无线信道进行类别划分和场景识别.

英文摘要：

In order to explore the propagative characteristics of wireless channel and establish corresponding wireless channel model,in this paper, in view of the ＂fingerprint＂ characteristics of wireless channel modeling and recognition problem for the scene, we put forward the modeling method of wireless channel which is based on Unscented Kalman Filter Neural Network（UKFNN）. Firstly, we use Karhunen-Loeve（K-L） to deal with channel transmission data,to realize the change of data from complex domain to real domain, and then use Principal Component Analysis（PCA） for data dimension reduction, after getting the dimension-reduced data, UKFNN is applied to build the characteristic model of wireless channel. Lastly, we establish the evaluation index of wireless channel according to the model-based output, and use the evaluation index to classify and recognize different scenes. The simulation results show that the model based on UKFNN can classify and recognize different scenes accurately.