关键词: 惯性-地磁组合, 自适应卡尔曼算法, 人体姿态跟踪, 估计残差

Abstract: There is a common problem existed in the existing attitude algorithm used for inertial-magnetic units, that is, some of those depend too much on the outputs of gyro, and have good dynamic but poor static performance, while the other of those depend too much on the outputs of Accelerometer and Geomagnetic sensor（AG）, and have good static but poor dynamic performance. It is problematic to use the module of the linear acceleration to regulate the dependence of gyro or AG, though it is sometimes effective. A method proposed in this paper is to estimate the module of the linear acceleration vector and the geomagnetic vector, and then takes the estimation error as the observation noise of the Extended Kalman Filter（EKF）, which is the most commonly used in the attitude algorithm, in order to construct an adaptive EKF. The estimation error is zero-mean and stationary, and most importantly, its variance increases significantly when the body is moving, so it can make the adaptive EKF has good static and dynamic performance. The experiments are constructed to compare the Kalman filter of MTi and ADIS16480 with this adaptive EKF, and the results validate the effectiveness of the proposed method.

Key words: inertial-magnetic units, adaptive Kalman filter, human body attitude tracking, estimation error