摘要
为了进一步改善非线性功率放大器系统的线性度,提出了一种基于BP神经网络逆向建模的离线训练自适应预失真方法。利用BP神经网络对功放逆向建模,并将建立好的逆模型参数作为预失真器模型初值。为了提高在初始预失真系统中预失真器的线性化效果及系统自适应进程的速度,在建立自适应预失真系统之前,利用BP逆向模型对预失真器进行离线训练。最后采用直接结构和最小均方(LMS)算法调节神经网络预失真器的权值,以消除放大器非线性的扰动。仿真结果显示,此方案可使邻道互调功率降低约18 dB,而经典的直接—非直接结构只降低了8 dB,表明此预失真方案能够更好地改善功率放大器的线性度。
In order to further improve the linearity of nonlinear power amplifier system, this paper proposed an adaptive pre- distortion method that was off-line training based on BP neural network inverse modeling. Firstly, it used BP neural network for inverse modeling of the amplifier, and took parameters of inverse model which had established as the initial of the predistortion model, then, in order to improve the effect of linearization of predistortion in the initial predistortion system, and accelerate the adaptation process of predistortion system, before establishing adaptive predistortion system, predistorter was off-line trained by BP inverse model. Finally, using direct structure and the LMS algorithm to adjust the weights of neural network pre- distorter, so as to eliminate nonlinear perturbations of amplifier. Simulation results show that this scheme can make adjacent channel intermodulation power reduce by about 18 dB, while the classical direct-indirect structure only reduced 8 dB. It indi- cates that this predistortion scheme can improve the linearity of the power amplifier better.
出处
《计算机应用研究》
CSCD
北大核心
2014年第4期1105-1108,共4页
Application Research of Computers
基金
国家自然科学基金资助项目(60971048)
辽宁省博士科研启动基金资助项目(20091033)
