Oracle approximating shrinkage estimator based cooperative spectrum sensing for dense cognitive small cell network

In this paper, we study the problem that dense cognitive small cells cooperate to sense primary signals of a macro cell. In consideration of the dense deployment of small cells, the number of small cells (sample dimension) is comparable to the number of sample (sample size), in which case sample covariance matrix is ill-conditioned estimator of high-dimensional statistical covariance matrix. The poor estimated performance of sample covariance matrix leads to degradation of sensing performance. Therefore, based on Neyman-Pearson theorem, we propose two oracle approximating shrinkage estimator based cooperative spectrum sensing (OAS-CSS) algorithms by utilizing oracle approximating shrinkage (OAS) estimator which is more accurate compared with sample covariance matrix. First method is proposed in case of ideal noise and we derive the theoretical expressions of probability of false and threshold. In the latter method, non-ideal noise is considered whose variances are imbalanced. Simulations show that our proposed OAS-CSS detectors exhibit better performance than traditional detectors and existing high-dimensional detectors. Also, theoretical sensing performance results with respect to ideal noise show an excellent agreement with simulation results when sample dimension and sample size are large.