Fusing Visuals with Magnetic Signals to Improve Indoor Localization Using Vision Transformer

Sensor fusion-based indoor localization is an evolving application that uses fused information to determine the location of smartphone users. However, the heterogeneity of sensors across various smartphones has significantly compromised the accuracy of localization algorithms. Therefore, this paper introduces MH-ViL, an infrastructure-free and calibration-free frame-work built on top of the Vision Transformer neural network. MH-ViL seamlessly integrates magnetic field signals (MFS) and visual images for localization tasks. A novel magnetic feature projection (MFP) model is proposed to effectively map MFS onto visual image features, enhancing positional accuracy within the self-attention mechanism. Real-time experiments demonstrate that MH-ViL surpasses alternative models, with an impressive 92% accuracy. We provide a 95% confidence interval with an error below 0.5 meters. Code: https://github.com/Hamaad1/MH-ViL.git