基于BP神经网络修正的自适应Singer模型

Adaptive Singer Model Based on BP Neural Network Modification

针对传统Singer模型在跟踪机动目标时存在稳态误差以及模型对加速度的先验统计量存在依赖等缺点,提出了基于BP神经网络修正的自适应Singer模型。该模型的样本数据选用卡尔曼滤波状态量中的加速度估计量,采用Burg算法估计加速度的功率谱密度,并利用BP神经网络对谱估计结果进行修正,进而导出当前统计模型下的Singer模型的参量。Simulink仿真结果表明,该模型能够克服传统Singer模型跟踪机动目标性能差的缺点,并且模型在收敛之后不再依赖加速度的先验统计量。

In order to solve the deficiency of Singer model in maneuvering target tracking, the Singer model of adaptive parameters is proposed. The model＇s steady error can be eliminated and it will not rely on the prior statistical information, The acceleration estimation in Kalman filtering is used, and Burg algorithm is employed to give the estimation of acceleration＇s PSD（Power Spectrum Density）. Finally, BP（Back Propagation） neural network is used to modify the Burg algorithm＇s estimation result. The Simulink simulation results are given that, the BP neural network modified model can overcome the deficiency of Singer model in maneuvering target tracking, and it will not rely on the prior statistical information after its convergence.