论文部分内容阅读
为了深入研究无线信道的传播特点,我们提出了无迹卡尔曼神经网络的建模方式,以此来解决移动网络通信中无限信道指纹特征建模问题以及场景的识别问题。首先,传播数据的预先处理,将所有数据由复数域转化为实数域,这其中会用到霍特林变换方法。其次处理数据的降维,利用主元分析得出降维后的数据,再利用无迹卡尔曼神经网络对移动互联网中的无线通信指纹特征进行模型的初步建立。最后根据输出数据建立无线信道的测评指标,根据场景不同对信道数据进行划分和识别。根据仿真数据可以得出,要想完成无线信道的类别划分以及场景的识别工作,采用无际卡尔曼神经网络建立一个完整的信道模型非常有必要。
In order to further study the characteristics of wireless channel propagation, we propose a modeling method of unscented Kalman neural network in order to solve the problem of infinite channel fingerprint feature modeling and scene identification in mobile network communication. First, the pre-processing of propagating data transforms all the data from the complex number domain to the real number domain, of which the Hotelling transform method is used. Secondly, the dimensionality reduction of the data is processed, the dimensionality reduction data is obtained by principal component analysis, and then the unshared Kalman neural network is used to establish the model of the wireless communication fingerprint feature in the mobile Internet. Finally, according to the output data to establish wireless channel evaluation index, according to the scene of the different channel data to be divided and identified. According to the simulation data, it can be concluded that it is necessary to establish a complete channel model by using the non-Invariant Kalman neural network in order to complete the classification of the wireless channels and identify the scene.