基于ε-支持向量回归机的盲均衡算法

Blind equalization algorithm based on ε-support vector regressor

在研究支持向量机的基础上,针对其优良的泛化性能,提出了基于ε-支持向量回归机(ε-SVR)的盲均衡算法。采用迭代加权二次规划方法进行求解,给出理论模型和算法推导,通过核函数推广到非线性盲均衡。计算机仿真结果表明,该算法在基带条件下对具有严重码间干扰的接收信号能够均衡处理,实现电话信道的零误差传输,相对于经典CMA算法而言,新算法在提高收敛性能的同时,提高收敛速度50%左右。尤其是在小样本情况下,计算复杂度低,收敛速度快。

Based on the support vector machine,a new blind equalization algorithm based one-support vector regressor （ε-SVR）was developed because of the good generalization properties the SVR presented. A iterative reweighed quadratic programming procedure was proposed for attaining the solutions, the theoretical model and algorithm derivation were also given, it can also be generalized to nonlinear blind equalization using kernel functions. Computer simulation results show that the algorithm equalizes received base band signal with serious inter-symbol interference and realizes error-free transmission in typical telephone channel. The proposed algorithm has better convergence performances than the classic constant modulus algorithm. The new algorithm improves the convergence performance, at the same time, it can improve the convergence at the rate of about 50%. Especially in small sample cases,the algorithm has lower computational complexity and faster convergence speed.