On the Eigenvalue-Based Spectrum Sensing and Secondary User Throughput

In this paper, we study the tradeoff between sensing time and achievable throughput of the secondary user that employs robust eigenvalue-based spectrum sensing techniques in the presence of noise uncertainty. First, we study exact distributions of the test statistics for two types of robust eigenvalue-based sensing techniques, namely, the blind generalized likelihood ratio test (B-GLRT) detection and energy with minimum eigenvalue (EME) detection. The developed threshold setting is more accurate than benchmark methods in achieving a target constant false alarm rate (CFAR). Second, prior to the throughput analysis, the necessary asymptotic detection and false alarm probabilities under noise uncertainty are formulated for eigenvalue-based detectors such as maximum eigenvalue detection (MED) and maximum-minimum eigenvalue (MME) detection. Finally, the throughput is maximized using eigenvalue-based spectrum sensing techniques which are B-GLRT, EME, MME, and MED detectors. The results are compared with the commonly used energy detector (ED). An improved achievable throughput is obtained under low-signal-to-noise-ratio (SNR) regime by incorporating the robust eigenvalue-based techniques, which are insusceptible to noise uncertainty.