Hybrid indoor localization using Wi-Fi–magnetic LSTM fingerprinting and vision-based distance fusion with EKF and particle filtering

Abstract

Accurate indoor localization remains challenging because Wi-Fi signals are affected by multipath propagation, signal fluctuation, and environmental changes, while vision-based methods may suffer from occlusion, perspective distortion, and lighting variation. This paper proposes a hybrid indoor localization framework that combines Wi-Fi Received Signal Strength Indicator (RSSI), magnetometer fingerprints, vision-based distance estimation, and probabilistic fusion to tackle these aforementioned challenges. The framework was evaluated in a third-floor indoor corridor environment at the Faculty of Engineering, Koya University, Iraq, using a 1 m grid consisting of 344 indoor reference positions. First, a Long Short-Term (LSTM)-based fingerprinting model estimates the initial indoor X, Y coordinate using Wi-Fi RSSI and magnetometer readings. Second, a vision-based model estimates the camera-person distance from Closed-Circuit-Television (CCTV) images using pose detection, depth-related features, geometric mapping, and residual correction. Finally, the LSTM coordinate and vision-derived distance are fused using Extended Kalman Filter (EKF) and particle-filter-based methods (PF), separately. The LSTM model achieved an exact-coordinate accuracy of 80.67% and an Root Mean Square Error (RMSE) of 1.5559. The vision model achieved an average distance error of 0.159 m. The fusion improved localization accuracy, reducing the error to 0.70 m using EKF and 0.40 m using the particle-filter-based method. The results show that vision-derived distance constraints can significantly improve Wi-Fi–magnetic fingerprint localization.