Abstract: To improve the tracking precision of Kalman filter online, the Gaussian degree of the random sequence is defined, which contains super-Gaussian, Gaussian, defective-Gaussian and non-Gaussian, it finds out the range that Kalman can work. In regard to the KF tracking precision is not well when the random sequence obeys the defective-
Gaussian, two parameters that can reflect the consistency of the predefined noise covariance and the real are mentioned. Adaptive parameter is proposed, which makes the predefined noise covariance fit the real, thus improves the tracking accuracy. The experimental results show that the tracking accuracy is well when the noise obeys super-Gaussian and Gaussian, yet bad when noise obeys defective-Gaussian and the predefined noise covariance is disagree with the real； while adaptive parameter is used, the tracking accuracy is improved greatly.

Key words: Kalman filter, Gaussian degree, adaptive parameter, tracking accuracy

摘要： 为了在线提高卡尔曼滤波算法（KF）的跟踪精度，对随机序列进行高斯度的定义，将随机序列的分布分为超高斯、高斯、次高斯和非高斯，找出KF可以工作的范围。针对噪声服从次高斯分布时，KF跟踪精度不高，介绍了两个可以判断KF使用过程中设定的噪声协方差与实际是否一致的参数。提出了参数自适应的方法，使设定的噪声协方差与实际值可以自适应地一致，从而提高了KF的跟踪精度。实验结果表明，噪声服从高斯和超高斯分布时，KF跟踪精度很高；噪声服从次高斯分布时，若噪声协方差的设定值与实际值不一致，跟踪误差会很大，对此使用了参数自适应法，可以大大提高KF的跟踪精度。

关键词: 卡尔曼滤波, 高斯度, 自适应参数, 跟踪精度