信道模型是无线通信系统设计和通信算法技术评估的基础，因此准确地对信道进行建模至关重要。目前主流的信道建模方法为基于几何统计的随机信道模型(Geometry-based Stochastic Channel Models, GSCM)，该模型具有复杂度折中、适用载频大、能反映信道的长期统计特性等特点。其建模核心思想是利用具有一定参数(大尺度参数、小尺度参数以及多普勒频率等)的簇的叠加来生成最终的信道矩阵。但是目前的GSCM信道模型是在二维平面内进行建模的，忽略了收发天线间的高度差，没有包含实际传播环境中俯仰维度的信息，缺乏对信道在三维空间内色散情况的建模，不能精确描述实际传播信道。因此本论文对从测量数据中提取3D信道参数、基于WINNER II模型建立完整的3D信道模型进行了研究。主要工作内容如下：1. 完整地提出了基于SAGE算法的3D信道参数提取方法。该算法可以实现使用均匀线性天线阵列(ULA)作为实际信道探测的天线设备，由实际探测得到的冲激响应(CIR)数据来提取信道中多径分量的空-时-频的三维信息。解决了SAGE算法在参数提取的实现中关于3D波程差的计算以及基于探测设备发射信号时间帧结构的发射脉冲信号建模的问题。算法所提取的多径参数可用于后期建模。本文给出了算法的实现结果，并对其性能进行分析。2. 在GSCM的建模方法和WINNER II信道模型的基础上，建立了完整的三维信道模型。引入了俯仰角信道空间参数，包括俯仰角度扩展、空间三维视距角以及移动台速度的俯仰参数，解决了三维波程差、三维多普勒频率以及天线阵列任意姿态带来的三维极化方向图变化对随机相位的影响的问题。首先通过链路内和链路间的参数相关性来生成包含俯仰维度的大尺度参数，然后以一定的数学分布随机生成小尺度参数，最后以这些参数构成每一条簇的信道冲激响应，通过将这些簇进行叠加建立3D信道模型。在所建立的3D信道模型中，用户可以自定义配置天线阵列的三维几何信息以及三维采样的场方向图，并通过极化增益分解的方法计算GCS坐标系下的天线响应增益。论文完整地生成了3D MIMO的信道系数矩阵，真实地模拟实际MIMO信道的传输，建立了一个通用3D信道模型。3. 基于所建立的3D信道模型，对3D信道的信道相关性以及信道特性进行仿真。论文提供了所建模信道的大小尺度参数统计特征曲线，以便于后期结合实测数据、数学统计建模理论以及不同仿真平台等因素对信道特征参数进行分析和修正，进而建立适合室内、室外及其它特殊应用场景下的3D无线信道模型。

Channel model is the basis for a wireless communication system design and communications technology assessment algorithm, and therefore it is very important to model the channel accurately. The mainstream channel model method is the geometry-based stochastic channel models (GSCM). It has a balanced complexity, a lager carrier frequency, a good performance in reflecting the channel long-term statistical properties, and a good adaptive ability in the realistic propagation environment. Its basic idea is superimpose the modeling clusters with certain parameters to generate a final channel matrix. As the 2D channel model cannot have a completely accurate description on the realistic propagation environment, this thesis proposes an algorithm which models a complete 3D channel based on GSCM model in WINNER II using the 3D channel parameters extracted from the measured data. The main contents of this thesis are as follows:1. Firstly, this thesis completely proposes an extraction method for SAGE-based 3D channel parameters. It achieves to use the uniform linear antenna array (ULA) as the realistic channel sounder. It uses the impulse response data obtained by detecting the realistic channel to extract the three-dimensional space-time-frequency information of multipath component. It also solves the 3D wave path-difference calculation problem in the parameter extraction of SAGE algorithm and the transmitted pulse signal model problem based on detection devices transmit signals time frame structure. The extracted multipath parameters can be used for modeling later. The experimental results and analyses about the proposed algorithm are given at last..2. Secondly, this thesis models a complete 3D channel model based on GSCM and WINNER II. It solves the problem that change of three-dimensional polarization pattern direction brings impact on the random phase by introducing channel spatial parameters on pitch angle. The change of three-dimensional polarization pattern direction can be influenced by the three-dimensional wave path-difference, the three-dimensional Doppler frequency and the arbitrary directions of antenna array. It produces the large-scale parameters based on the correlation links within and between the links. The small-scale parameters can be generated randomly by the large-scale parameters based on a certain mathematical distribution. The channel impulse response of each rays are constituted by these parameters. A final 3D channel model is modeled as a cluster comprising a superposition of these rays. Users can customize to configure the three-dimensional geometry information of antenna array and the three-dimensional sampled field pattern. The antenna response gain of GCS coordinate can be calculated using polarization gain decomposition method. Finally, a general channel model can be generated by using a complete 3D MIMO channel coefficient matrix to simulate the actual MIMO channel transmission realistically.3. Finally, the related Matlab programs are written to extract channel parameters and establish channel model. In this thesis, the experimental results about channel correlation, channel characteristics and parameter statistical characteristic curves of the proposed 3D channel model have been provided. It will be helpful for expediently analyzing and correcting the channel characteristic parameters by binding measured data, mathematical statistical modeling theory and different simulation platform later in order to establish domestic radio channel model which conforms to the realistic propagation environment in indoor, outdoor and other special application scenarios.