Mean Spectral Radius Detection for Cognitive Radio

In this paper, a new spectrum sensing algorithm is proposed based on the eigenvalue distribution of the covariance matrix of sensing nodes. The received signals of all the nodes can be denoted by a non- Hermitian random matrix. A recent research indicates that the eigenvalue distribution for the product of non-Hermitian random matrices follows Single Ring Theorem for the noise-only case. However, for the signal-present case, the inner radius of the eigenvalue distribution is smaller than that of the noise-only case. Then mean spectral radius (MSR) can be utilized to detect the signal. The proposed method overcomes the noise uncertainty and has higher detection performance than the maximum- minimum eigenvalue (MME) detection when the primary signals among sensing nodes are uncorrelated. Finally, Simulations are performed to verify the effectiveness of the proposed method.