A Practical Hybrid Height Estimation Algorithm Based on FM-Aided Motion Mode Recognition

Abstract

Height estimation is critical for precise indoor localization in multistorey buildings. Traditionally, it is achieved by the barometric altimetry (BA)-based method, whose accuracy suffers from ambient barometric pressure fluctuations. Zero-velocity-update (ZUPT)-aided acceleration integration is another applicable approach, where zero velocity moments are introduced as calibration reference to limit error accumulation. However, zero velocity moments are derived from foot-mounted sensors in the stairs-climbing scenario, making it unavailable for other sensor locations and scenarios such as elevators. In this article, we propose a hybrid height estimation algorithm with both high accuracy and improved practicability. With the assistance of motion mode recognition (MMR), we apply a constraint on height estimations from both the BA-based method and the integration-based method, and then produce our height estimation by fusing the constraint heights. First, four typical motion modes comprising standing still, plane walking, climbing stairs, and taking elevator are recognized with features from acceleration, barometric pressure, and frequency-modulated (FM) signal. Second, we constrain the height estimations based on MMR results. To improve the BA-based method, a reference pressure update strategy is proposed to make the method applicable in an environment without reference barometers. Besides, the step-based integration strategy and the reverse compensation strategy are designed to optimize integration-based heights in stairs and elevators scenarios, respectively. Finally, the height estimations from these two sources are fused adaptively. Experiment results show that the accuracy of the proposed hybrid method outperforms that of the BA-based method. Moreover, the introduction of FM signal features effectively improves the performance of MMR and, thus, height estimation.