Paper Info
Title
Deep Neural Network-Based Robust Spectrum Sensing: Exploiting Phase Difference Distribution
Abstract
As an enabling technology to address spectrum shortage, spectrum sensing has been investigated a lot. However, the uncertainties in the detection environment, including noise uncertainty and carrier frequency (CF) mismatch, still remain as the main challenges of spectrum sensing, which greatly degrades the sensing performance of typical sensing methods, such as energy detection and cyclostationary detection. To this end, this paper proposes two robust spectrum sensing schemes by leveraging the difference between the phase difference (PD) distribution of noise-perturbed signal and that of Gaussian noise. Specifically, the compact approximation of the PD distribution is first derived to enable the extraction of the features of PD distributions, which are robust to noise uncertainty and CF mismatch. Based on these features, two sensing schemes based on the deep neural network (DNN), referred to as DNN-based PD distribution detection (PDD) and blind PDD (BPDD), are proposed to detect spectrum holes in cases with known CF and unknown CF, respectively. Simulation results show that our proposed schemes are more robust to CF mismatch and noise uncertainty in comparison with the existing sensing schemes. Furthermore, when the CF of the sensed signal is unknown, the proposed BPDD significantly outperforms existing blind sensing schemes.
Year
DOI
Venue
2021
10.1109/ICC42927.2021.9500743
IEEE INTERNATIONAL CONFERENCE ON COMMUNICATIONS (ICC 2021)
Keywords
DocType
ISSN
Robust spectrum sensing, deep neural networks, phase difference distribution
Conference
1550-3607
Citations 
PageRank 
References 
0
0.34
0
Authors
7
Name
Order
Citations
PageRank
Yang Wang111.04
Wenjun Xu24511.81
Zhijin Qin322.06
Yimeng Zhang411.70
Hui Gao512716.33
Miao Pan65716.43
Jiaru Lin764680.74