Joint DOA and Polarization Estimation Based on Tensor Train Decomposition and Coupled Tensor Decomposition for Cylindrical Vector Conformal Array

This paper investigates the performance degradation of joint direction-of-arrival (DOA) and polarization estimation for vector conformal arrays (VCA) under low signal-to-noise ratio (SNR) and limited snapshot conditions. To address this problem, two model driven tensor based DOA–polarization estimation algorithms are proposed for cylindrical VCA: tensor train rank-1 singular value decomposition multiple signal classification (TTr1 SVD-MUSIC) and coupled canonical polyadic decomposition multiple signal classification (C-CPD-MUSIC). The TTr1-SVD-MUSIC algorithm introduces tensor train rank-1 decomposition into VCA based DOA estimation by directly applying cyclically unfolded TTr1-SVD to a single tensor signal model. This enables hierarchical low-rank modeling and direct signal subspace extraction without forming high-dimensional covariance matrices, thereby significantly reducing computational complexity while preserving multidimensional signal structure. In addition, the C-CPD-MUSIC algorithm establishes a coupled tensor signal model by explicitly exploiting the inherent horizontal–vertical spatial coupling of cylindrical VCA. Through coupled tensor factorization with shared factor matrices, the signal subspace is jointly reconstructed across multiple spatial dimensions. Both algorithms integrate the MUSIC criterion to achieve automatic pairing of DOA and polarization parameters. Simulation results demonstrate that C-CPD-MUSIC achieves superior accuracy and robustness under low-SNR and limited-snapshot conditions, while TTr1-SVD-MUSIC provides a favorable trade-off between estimation performance and computational efficiency.