Ultrawideband Non-Line-of-Sight Classification Using Transformer-Convolutional Neural Networks

Ultrawideband technology enables high-precision indoor positioning, but its performance degrades significantly in non-line-of-sight environments due to signal obstruction. Accurate identification of non-line-of-sight channels is essential for mitigating localization errors. While recent deep learning methods, such as convolutional neural networks and long short-term memory models, have been applied to classify non-line-of-sight scenarios from raw channel impulse response data, they often struggle with capturing long-range dependencies, efficiently extracting spatial and temporal features, resulting in suboptimal classification performance. This paper proposes a novel hybrid model combining Transformer architecture and convolutional neural networks for non-line-of-sight classification using raw channel impulse response data from received ultrawideband signals. The Transformer captures global temporal dependencies through self-attention, while convolutional neural networks efficiently extract local spatial features. Experimental results on a public dataset show that the proposed model outperforms benchmark methods in widely used evaluation metrics, highlighting its effectiveness in improving non-line-of-sight detection for ultrawideband-based indoor positioning.