Modeling with noises for inertial sensors

ARMA-modeling the inertial sensor's colored noise must determine the AR parameters, the MA parameters, as well as the variance of the measurement white noise when the order of the ARMA model is estimated. Due to the existence of the MA items, the Yule-Walker equation constructed by the colored noise's autocovariances starts from the order higher than the order of MA or AR model, which prevents from further improvement of the ARMA-modeling accuracy. In this paper, the ARMA model is approximated to a high-order AR model. Since there are no MA items in the approximated AR model, the Yule-Walker equation can be constructed from the 1st-order of the colored noise's autocovariances, which is beneficial to improving the estimation accuracy of the white noise variance. This method can also be used to estimate the AR parameters accurately. Simulations and experiment validate the effectiveness of the method. The length of the colored noise used in the ARMA modeling is also determined quantitatively.