Cellular Data-Based Indoor Localization for Smart Health Applications

Indoor localization is a crucial component of modern smart environments, supporting applications in healthcare, security, and asset tracking. Traditional methods, such as Wi-Fi fingerprinting and Bluetooth-based tracking, require extensive calibration and infrastructure, limiting their scalability. In this study, we present an LTE-based indoor localization framework that leverages key LTE performance indicators to predict grid locations within an indoor environment. Using a multi-class classification approach, our model achieved an accuracy of 95.78% by analyzing signal strength variations and propagation characteristics. We evaluated classifier performance using precision-recall curves, F1-scores, ROC analysis, and feature space visualization techniques like PCA and t-SNE. Results indicate that LTE KPI-based localization effectively distinguishes between locations, although classification challenges persist in adjacent regions with signal overlap. Further analysis suggests that environmental factors, class imbalance, and temporal variations contribute to misclassification in certain areas. To address these challenges, future work will incorporate temporal dependencies, feature augmentation, and advanced post-processing techniques to refine classification accuracy. Our findings demonstrate that LTEbased localization offers a scalable and infrastructure-independent alternative to conventional methods, enabling real-time tracking with minimal deployment effort.