Novel Estimation of the Number of Sources in Radar Signals via Sparse Representation
Seung-Jae Lee, Kwan-Young Oh · IEEE Access · 2026
In this letter, we propose a novel method to estimate the number of sources in radar signals via sparse representation. Assuming doublets as in ESPRIT, the eigenvector matrix of computed from the covariance matrix of the received radar signals is first divided into two sub-eigenvector matrices. Then, we apply two important assumptions, considering that 1) the signal subspaces in the two sub-eigenvector matrices can be represented sparsely, and 2) that the noise subspaces are densely representable. Given those constraints, the number of sources can be determined by a sparse recovery algorithm based on compressive sensing. The proposed approach achieved better performance than existing methods and remained stable in various conditions. In particular, it achieved robust performance even when sources were located at very close range or the SNR values were low.