On the use of eigenvectors for signal detection and classification in multiple antenna cognitive radios

This paper presents a comparative analysis of eigenvector-based signal detection algorithms and a new eigenvector-based modulation classifier for multiple antenna cognitive radio applications. The signal detection algorithms considered in this work are: Feature Template Matching (FTM), Multiple Feature Matching (MFM) and Principal Components (PC)-based detection. A software-defined radio (SDR) platform is used to experimentally evaluate the detection performance of these algorithms. It is observed that MFM, which uses four signal features extracted across space, time and phase dimensions, provides the best performance among the considered algorithms. This paper also presents a new idea on the use of eigenvectors for signal classification in multiple antenna systems. Using computer simulations, it is shown that eigenvectors corresponding to different modulation schemes allow for reliable signal identification even at very low SNR values (on the order of -20 dB), therefore they can be used effectively for signal classification in cognitive radio applications.