A New Method for Source Number Estimation in the Presence of Unknown Nonuniform Noise

Classical source number estimators are usually derived under the assumption of uniform white noise; hence, they are ineffective with unknown nonuniform noise. Current advanced estimators designed for nonuniform noise are computationally expensive and often unable to yield satisfactory performance in unfavorable conditions, such as low signal-to-noise ratio, small number of snapshots, and sources with different transmit power. To address this, this paper proposes a new method to estimate the number of sources under nonuniform noise. The proposed method first constructs a likelihood ratio statistic as a function of the maximum likelihood estimation of the array covariance matrix, which is estimated by a subspace estimation algorithm. Then, based on the asymptotic theory of the likelihood ratio, the number of sources is estimated via a sequence of hypothesis tests. Theoretical analysis demonstrates that the proposed estimator is consistent in the general asymptotic regime. Simulation results show that the proposed estimator achieves a higher correct detection probability in unfavorable conditions and is more robust against nonuniformity of noise than state-of-the-art estimators.