提出一种基于粒子滤波的非线性非高斯信号分离方法.该方法依据状态空间模型把信号分离问题转化为信号的状态和参数的联合估计问题,利用粒子滤波方法,结合核平滑收缩技术拟合系统未知参数后验分布,以实现非线性系统中多路信号的分离.仿真结果表明,与现有分离算法相比,该方法能有效解决非线性非高斯系统中多路信号的分离问题,并提高未知参数的估计精度.
In this paper,a novel separation method for nonlinear non-Gaussian signal is proposed.According to the state space model,the issue of separation was changed into the joint estimation of the parameter and state of the signal.In order to solve the problem of multi-signal separation for nonlinear systems,the algorithm uses particle filter method,combined the kernel smoothing contraction technique.The simulation results show that,in nonlinear system,compared with DEKF algorithm,the proposed method can solve the problem of multi-signal separation effectively.