DS2MA: A Deep Learning-Based Spectrum Sensing Scheme for a Multi-Antenna Receiver
Keunhong Chae, Yusung Kim
Abstract
In this letter, we propose a novel deep learning-based spectrum sensing scheme using a multi-antenna receiver. Our main idea is constructing a correlation matrix composed of not only auto-correlation functions per each antenna but also cross-correlation functions between antennas. By using the rich informative matrix, with a simple convolutional neural network (CNN) structure, our model, DS2MA (Deep Spectrum Sensing with Multiple Antennas), can efficiently learn to detect the presence of a primary user (PU). In our extensive simulation results, DS2MA using only auto-correlation functions for each antenna can outperform previous related works. By adding cross-correlation functions together, DS2MA can significantly improve the detection performance. We also show the impact of impulsive noise and correlation coefficient. The simulation results show that the proposed DS2MA provides a higher detection probability compared to the existing works regardless of the impulsive noise impact and correlation coefficient.