收发隔离是干扰机系统中的关键问题,若隔离度不够,轻则削弱干扰机效率,重则造成自激使干扰机不能正常工作。传统的最小均方(LMS)算法应用到基于自适应系统辨识的收发隔离中时,由于精度不够,隔离性能并不理想。针对该问题,提出一种分段变步长归一化LMS(NLMS)算法。该算法利用迭代系数状态因子对迭代系数与真实系数逼近状态进行分段,当迭代系数状态因子处于不同的区间时,采用不同的变步长方案。理论分析和仿真证明:与基于其他LMS 算法的隔离方案相比,基于分段变步长NLMS 算法的收发隔离方案隔离性能有较大的改善。

Transmitter-receiver isolation is an important problem in jamming system. Low efficiency of jammer and even self-excitation can be caused when the degree of transmitter receiver isolation is low. Transmitter-receiver isolation is not good when traditional LMS algorithm is implemented because of worse convergence accuracy. To solve this problem, a segmental variable step-size normalized LMS algorithm based on the state parameter of iteration coefficient is proposed. The state parameter of iteration coefficient is employed in this normalized LMS algorithm to express the approximation degree between iteration coefficient and real coefficient. Different kinds of variable step-size scheme are adopted when the value of state parameter of iteration coefficient is located in different segment. The theory analysis and simulation results demonstrate that the performance of transmitter-receiver scheme based on the proposed segmental variable step-size normalized LMS algorithm is better than schemes based on other LMS algorithms.