GLRT-Based Spectrum Sensing With Blindly Learned Feature Under Rank-1 Assumption
Prior knowledge can improve the performance of spectrum sensing. Instead of using universal features as prior knowledge, the authors propose to blindly learn the localized feature at the secondary user. Motivated by pattern recognition in machine learning, they define signal feature as the leading eigenvector of the signal's sample covariance matrix. Feature Learning Algorithm (FLA) for blind feature learning and Feature Template Matching algorithm (FTM) for spectrum sensing are proposed. Furthermore, they implement the FLA and FTM in hardware. Simulations and hardware experiments show that signal feature can be learned blindly. In addition, by using signal feature as prior knowledge, the detection performance can be improved by about 2 dB.