Expectation–maximization-based Kalman filter under colored measurement noise for INS-based integrated human localization

An increasing number of fields are using precise location these days. However, colored measurement noise (CMN) can affect the localization accuracy of data-fusion filters. The aim of this research is to present an adaptive Kalman filter (KF) that employs the approach of expectation maximization (EM) within a CMN framework for integrated human localization based on inertial navigation systems (INSs). Herein, an INS-based integrated model under CMN is derived, which employs the backward Euler method to reduce the influence of CMN. In this model, we use EM to enhance the accuracy of estimating noise statistics for KFs under CMN (cKFs). Further, an adaptive strategy based on the Mahalanobis distance is proposed, which can render KFs with high adaptability under actual positioning environments. The results of two real-world tests indicate that the proposed adaptive cEM-KF (cEM-KF: EM-based KF under CMN) outperforms the conventional KF, cKF, and cEM-KF with regard to position localization.