Description of Invention
Fingerprint-based indoor localization provides an effective solution for GPS-denied environments with minimal hardware requirements. Despite their popularity, these systems often suffer from signal fluctuations caused by shadowing, fading, and multipath effects, which undermine positioning accuracy. While 1D and 2D CNNs are capable of extracting spatial features, they do not account for temporal variations, impacting online localization accuracy. Although 3D CNNs can extract spatio-temporal information, their high computational demands hinder real-time applications. This study proposes a novel 3D-separable CNN that employs depth-wise and point-wise convolutions, complemented by skip connections and group shuffling. This architecture aims to address the computational constraints while preserving accuracy, making it suitable for real-time localization tasks.