基于边际似然比的目标机动的序贯检测方法

Marginal Probability Ratio Based on Method of Maneuver Sequential Test

总结了两大类机动检测的统计检验方法:χ2统计量检验法和似然函数比检验法。将一种基于边际似然函数比的序贯检测方法(Sequential Marginal Likelihood Ratio Test,SMLRT)应用到目标机动检测中,利用机动假设的先验分布推导出机动似然比的近似公式。实验比较了IMM(Interacting Multiple Model,IMM)、MR(Measurement Residual,MR)、IE(Input Estimation,IE)以及SMLRT在不同机动方式下的机动检测性能,数据表明:当假设的机动加速度变量的先验分布(混合高斯分布)与实际分布较为接近时,SMLRT的机动检测性能优于其他三种检测器的性能。

This paper generalizes two classes of statistical maneuver test methods： χ2 statistic test and probability ratio test.We apply a method of sequential marginal likelihood ratio test（ SMLRT） to maneuver detection. We compare the performance of this proposed method with IMM,MR and IE＇s. Simulation shows that SMLRT has its priority comparing others＇,when the prior distribution of maneuver acceleration fits（ Mixture Gaussian） to the real distribution.