.png){kind=icon}
.png){kind=icon}
Return
Efficient Spectrum Sensing via a Multi-Scale Dual-Path Segmentation Network
.png){kind=icon}
PRE
AI
DOI：10.1109/LWC.2025.3564151.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This letter introduces an advanced spectrum sensing (SS) method for next-generation wireless networks that utilize dual-path architecture called DPSegNet. This innovative model is designed to precisely segment 5G new radio (NR) and long-term evolution (LTE) signals by identifying the spectral content based on the frequency and time occupied by the signals. DPSegNet incorporates a context path to capture high-level semantic information, a spatial path to preserve detailed spatial features, and a novel feature fusion mechanism to effectively combine information from both pathways. This architecture effectively learns both local and global spectral features, thereby significantly enhancing its segmentation performance. The experimental results highlight the efficiency and effectiveness of DPSegNet, with a compact architecture consisting only 7M parameters, it achieves high performance with a global accuracy of 97.25% and a mean intersection-over-union (IoU) of 94.76%. These results demonstrate that DPSegNet is a highly promising solution for next-generation wireless communication systems.
Keywords:
5G new radio
cognitive radio systems
spectrum sensing
signal identification
deep learning
image segmentation
Journal
IF:
11.5
Papers:
2.7K
Citations:
1.3W