基于仿射投影算法的AR模型参数估计方法

The AR model parameters estimation approach based on affine projection algorithm

摘要:

为取得更有效的预报效果,在深入分析传统LMS(Least mean square)算法的基础上,提出利用仿射投影算法对AR模型参数进行自适应估计,利用实测的动态数据结合AIC(Akaike information criterion)准则建立自适应AR(Autoregressive)预报模型,并将该模型应用于船舶运动预报中,实例仿真比较分析表明:相对于LMS算法、LMS-Newton法和NLMS(归一化LMS)算法,基于仿射投影算法得到的AR预报模型,预报精度更高、预报时间更长,且在自适应AR模型参数估计中具有更快的收敛速度,能为实时在线预报提供理论依据.

Abstract:

To obtain more effective prediction effects,based on study of LMS algorithm,affine projection algorithm is proposed in this paper which is apply to estimating AR model parameters.The adaptive AR predicted model is built by using measured dynamic data combining with AIC criterion and the model is applied in the ship motion prediction.After comparative analysis of the simulation results,the predicted model based on Affine projection algorithm can get more precise prediction and the prediction time is more longer.The convergence rate of adaptive estimation of parameters about AR model is faster compared with LMS algorithm,LMS-Newton algorithm and NLMS algorithm.So the method provides theoretical basis for real time on-line prediction of ship motion.