A Non-Cooperative Transmitter Localization Algorithm Based on Deep Learning with Time-Frequency and Spatial Angular Information

The precise localization of non-cooperative transmitters is crucial for security applications and daily life. However, traditional geometry-based positioning techniques suffer from large errors due to multipath effects and nonline-of-sight (NLOS) propagation. We propose a novel non-cooperative transmitter localization algorithm that leverages deep learning to overcome these limitations. Inspired by fingerprint-based positioning, our approach utilizes readily obtainable information at the receiving end, including spectrum sensing, power spectrum analysis, and spatial angle information obtained through beamforming. By integrating this information, the algorithm effectively utilizes NLOS paths, often discarded in conventional methods, to achieve more accurate source localization. Our proposed system demonstrates a positioning accuracy of approximately 2.4 <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$m$</tex>.