Low-complexity Implementation Method for GLRT Detector for Rank-one Signal in Cognitive Radio

The generalized likelihood ratio test (GLRT) detector has excellent performance in detecting rank-one signal. However, the high computational complexity hinders its further application. To this end, a low-complexity implementation method based on the numerical analysis theory for the detector is proposed. Our method uses the accelerated power method based on Rayleigh quotient to iteratively compute the maximum eigenvalue, which has a faster convergence rate in detecting high-dimensional signals compared with the traditional power method; meanwhile, a direct threshold calculation method based on the cubic spline interpolation theory is proposed, which can quickly determine the threshold corresponding to any given target false-alarm probability. Our method effectively improves the computational efficiency and reduces the implementation complexity of the GLRT detector for rank-one signal. Finally, the simulation results demonstrate the effectiveness of the proposed method.