基于LS-SVM的宽带接收前端非线性补偿算法  

Nonlinearity Mitigation Method Based on LS-SVM for Wide-Band Receiver

作　　者：黄家露 王文涛[1] 周莲[1] 李姝[1] 杨波[1] 杨阳[1] 刘昭涛 高星寒 宋海平[1] HUANG Jia-lu;WANG Wen-tao;ZHOU Lian;LI Shu;YANG Bo;YANG Yang;LIU Zhao-tao;GAO Xing-han;SONG Hai-ping(Department of Information and Control,China North Vehicle Research Institute,Beijing 100072,China)

机构地区：[1]中国北方车辆研究所信息与控制技术部,北京100072

出　　处：《电子学报》2023年第6期1500-1509,共10页Acta Electronica Sinica

摘　　要：针对目前常用的基于参数化非线性模型(Parameterized Nonlinear Model,PNM)的补偿算法存在易陷入局部最小值,导致补偿性能不稳的问题,该文提出了基于最小二乘支持向量机(Least Squares Support Vector Machine,LS-SVM)的宽带接收前端非线性补偿算法.该算法基于减谱-时频变换法(Spectrum Reduction Algorithm based on Time-Frequency Conversion,SRA-TFC)盲分离接收前端输出信号中的大功率基波信号和其他小功率信号,并以此作为LS-SVM逆模型的训练输入-输出样本对.引入最小二乘支持向量回归(Least Squares Support Vector Regression,LS-SVR)算法高精度拟合接收前端非线性逆模型.通过以宽带接收前端的输出信号为测试样本消除其非线性失真分量.仿真与实测结果表明:该算法可使宽带接收前端的无杂散失真动态范围(Spurs-Free-Dynamic-Range,SFDR)提高约20 dB,较基于PNM的补偿算法提高了约5 dB.To address the problem that the commonly used compensation algorithms based on parametric nonlinear model(PNM)are prone to fall into local minima,leading to unstable compensation performance,a nonlinear compensation algorithm for broadband receive front ends based on least squares support vector machine(LS-SVM)is proposed.The algo⁃rithm blindly extracts the high-power fundamental signal and other low-power signals from the receiver output signal based on the reduced-spectrum-time-frequency transform(SRA-TFC)method,and use them as the training input-output sample pairs of the LS-SVM inverse model.The inverse model is then fitted with high accuracy by least squares support vector re⁃gression(LS-SVR)algorithm.The output signal of the wideband receiver is used as the test sample to eliminate its nonlin⁃ear distortion components.The simulation and measurement results display that the algorithm can improve the spurious free dynamic range(SFDR)of the wideband receiver by about 20 dB and it is increased by 5 dB compared with those meth⁃ods based on PNM.

关 键 词：宽带接收前端 非线性补偿 最小二乘支持向量机 最小二乘支持向量回归算法 无杂散失真动态范围 

分 类 号：TN92[电子电信—通信与信息系统]