Large Array Antenna Spectrum Sensing in Cognitive Radio Networks

We investigate spectrum sensing in cognitive radios with large antenna arrays. Gigantic MIMO and extra-large array antennas are fundamental technologies in 5G and future 6G networks, as well as other wireless networks like Wi-Fi. In conventional multiple antenna spectrum sensing, the likelihood ratio test is standard when samples outnumber antennas. It relies on covariance matrix estimates from the sample covariance matrix and its eigenvalues. However, in large array contexts where the number of antennas matches or exceeds the number of samples, the likelihood ratio test falters due to challenges in accurately estimating the covariance matrix from the sample covariance matrix. To address this limitation, we propose and derive three detectors based on linear spectral statistics from random matrix theory: the generalized likelihood ratio, Frobenius norm, and Rao tests, similar to conventional multiple antenna spectrum sensing. We compute the Gaussian distribution of these detectors under the noise-only hypothesis using the central limit theorem for high-dimensional random matrices. Performance evaluation examines how antenna and sample numbers influence outcomes, as confirmed by simulation outcomes. Our results illustrate the effectiveness of the proposed detectors in various real-world situations, encompassing correlated or uncalibrated antennas in networks featuring large array antennas. These detectors exhibit notable benefits, including decreased average signal-to-noise ratio and enhanced detection probability. Their versatility extends to multiple domains, including gigantic MIMO wireless communications, radar systems, and astronomical observations.