论文部分内容阅读
最大似然估计是提取目标微动特征参数的最佳估计方法,但直接用网格法求解计算量巨大,且激光探测微多普勒回波信号对应的代价函数具有高度非线性,存在多个局部最大值。为此,提出均值似然估计与蒙特卡罗结合的估计方法,给出了最大似然参数估计的闭合表达式,再通过设计压缩似然函数获得全局最大值,通过蒙特卡罗法抽样并计算循环均值估计出参数。该方法避免了传统方法中对高精度初始值和复杂迭代算法的依赖,能够实现参数的联合估计。对于多分量微多普勒信号,该方法可在参数估计的同时实现各微动分量分离,且不增加算法的复杂性。对仿真和实验数据进行估计,结果表明,该方法在达到近似于最大似然估计性能的同时可有效降低计算复杂度并确保了全局收敛,实现信号的分离和参数估计。
The maximum likelihood estimation is the best method to extract the inversed parameters of the target. However, the computational cost is huge when using the grid method directly. And the cost function of the micro-doppler echo detected by the laser is highly nonlinear, Local maximum. For this reason, the method of combining the mean likelihood estimation with the Monte Carlo method is proposed, the closed expression of the maximum likelihood parameter estimation is given, and the global maximum is obtained by designing the compression likelihood function, which is sampled and calculated by Monte Carlo method The mean value of the cycle estimates the parameter. This method avoids the dependence of high-precision initial value and complex iterative algorithm in the traditional method and can achieve the joint estimation of parameters. For multi-component micro-Doppler signals, this method can realize the separation of each micro-component while estimating the parameters without increasing the complexity of the algorithm. The simulation and experimental data are estimated. The results show that this method can reduce the computational complexity and ensure the global convergence while achieving the approximate maximum likelihood estimation performance, and realize the signal separation and parameter estimation.