Efficient CNN-Based Classification of NLOS/LOS Conditions Using CIR Features in Ultra-Wideband Networks

Indoor Positioning Systems (IPS) are crucial for location-based services and IoT. Ultra-Wideband (UWB) offers high precision but faces challenges from Non-Line-of-Sight (NLOS) conditions, which degrade accuracy. This paper introduces a machine learning model based on a Convolutional Neural Network (CNN) for NLOS vs. Line-of-Sight (LOS) classification in UWB systems. The proposed model operates by extracting relevant features from the Channel Impulse Response (CIR), rather than using CIR samples as in previous work, leading to a reduced computational complexity while maintaining high accuracy. The model was evaluated on an open source UWB dataset, showing a high classification performance and highlighting the potential role of deep learning in enhancing NLOS detection and potentially improving the reliability of positioning systems.