论文部分内容阅读
针对功放线性化算法中常用的Hammerstein模型,本文提出了一种基于信号高阶统计量的两步辨识方法,分别提取其线性记忆模块及无记忆非线性模块的参数。首先利用输出信号的高阶累积量的特殊切片,构建Hankel矩阵并将模型记忆深度的确定转换为矩阵的求秩问题,并同时提取线性记忆模块的参数,理论推导表明,模型的非线性效应并不影响线性效应的辨识;提出了一种迭代算法以提取无记忆非线性模型的参数,结果表明,若无记忆非线性模块传函为奇函数时,利用具有对称分布的独立同分布信号(I.I.D)作为激励,仅需一次迭代即可以求得非线性系数的全局最优值。仿真结果验证了该方法的可行性和高效性。
For the Hammerstein model, which is commonly used in the linearization algorithm of power amplifier, a two-step identification method based on high-order signal statistics is proposed in this paper. The parameters of linear memory module and non-memory nonlinear module are extracted respectively. Firstly, the Hankel matrix is constructed by using special slices of higher-order cumulants of the output signal, and the determination of model memory depth is transformed into the rank-seeking problem of matrix. At the same time, the parameters of linear memory module are extracted. The theoretical derivation shows that the nonlinear effect of the model The proposed algorithm does not affect the identification of linear effects. An iterative algorithm is proposed to extract the parameters of the memoryless nonlinear model. The results show that if the memorizer of memoryless nonlinearity module is an odd function, an independent and identically distributed IID ) As a stimulus, the global optimum of the nonlinear coefficients can be found in only one iteration. Simulation results verify the feasibility and efficiency of this method.