TY - GEN
T1 - Real-World IMU-Wi-Fi Fusion for Indoor Localization Using Deep Learning and Scale-Factor-Adaptive Kalman Filter
AU - Kong, Yuzhuo
AU - She, Changyang
AU - Wei, Xing
AU - Zhong, Yi
AU - Hassan, Syed Ali
N1 - Publisher Copyright:
© 2026 IEEE.
PY - 2026
Y1 - 2026
N2 - This paper presents a hybrid indoor localization framework that fuses Wi-Fi and inertial measurement unit (IMU) sensing on commodity devices. We designed a convolutional neural network and a long short-term memory network to extract motion increments from IMU measurements, while a graph neural network is used to refine Wi-Fi-based location based on round-trip time (RTT) and received signal strength (RSS). The location and the corresponding increment are integrated through a scale-factor-adaptive Kalman filter to achieve high-accuracy trajectory estimation. Real-world experiments in mixed line-of-sight (LoS) and non-line-of-sight (NLoS) environments demonstrate that the proposed system achieves sub-meter localization accuracy and is more robust than baselines, including Wi-Fi only method, IMU only method, conventional Kalman filter (KF) and dead reckoning algorithm.
AB - This paper presents a hybrid indoor localization framework that fuses Wi-Fi and inertial measurement unit (IMU) sensing on commodity devices. We designed a convolutional neural network and a long short-term memory network to extract motion increments from IMU measurements, while a graph neural network is used to refine Wi-Fi-based location based on round-trip time (RTT) and received signal strength (RSS). The location and the corresponding increment are integrated through a scale-factor-adaptive Kalman filter to achieve high-accuracy trajectory estimation. Real-world experiments in mixed line-of-sight (LoS) and non-line-of-sight (NLoS) environments demonstrate that the proposed system achieves sub-meter localization accuracy and is more robust than baselines, including Wi-Fi only method, IMU only method, conventional Kalman filter (KF) and dead reckoning algorithm.
KW - Deep Learning-Based Inertial Odometry
KW - Graph Neural Network
KW - Indoor localization
KW - Scale-Factor-Adaptive Kalman Filter
UR - https://www.scopus.com/pages/publications/105037333734
U2 - 10.1109/EMC68537.2026.11441953
DO - 10.1109/EMC68537.2026.11441953
M3 - 会议稿件
AN - SCOPUS:105037333734
T3 - Proceedings of 2026 International Conference on Embedded Systems, Mobile Communication and Computing, EMC2 2026
SP - 357
EP - 363
BT - Proceedings of 2026 International Conference on Embedded Systems, Mobile Communication and Computing, EMC2 2026
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2026 International Conference on Embedded Systems, Mobile Communication and Computing, EMC2 2026
Y2 - 12 January 2026 through 14 January 2026
ER -