连续型野值判别的新方法

New judgment method of continuous outliers

针对连续型野值严重影响卡尔曼滤波器可靠性这一情况,提出一种新的野值判别方法,即正交模型参数变化判别法.当新息序列满足以下3个条件即信号是有穷均值,观测噪声是均值为零的独立分布过程,信号与噪声互不相关时,卡尔曼新息序列估计值具有正交性质.正交模型参数变化判别法的实质是通过判断卡尔曼滤波新息序列估计值的正交性质是否丧失而判别出观测值中是否存在野值.通过仿真结果表明,新的方法对于连续型野值的判别方便易行,快捷有效.

In view of the problem that reliability of Kalman Filtering would be affected if there were outliers in the measurement values, a new judgment method is proposed, which is called orthogonality judgment method. When innovation sequence is satisfied with three conditions i.e. signals are finite mean values; measurement noises are independent distribution of Zero Mean Values; and signals and measurement noises are incorrelate with each other, the estimated values of Kalman innovation sequence have orthogonality character. The nature of orthogonality judgment method is to judge if the orthogonal of innovation values of Kalman Filtering is lose or not. The simulation results show that the new method is simple, availability and swift to make out continuous outliers.