A novel joint two-dimensional angle and frequency estimation method based on a uniform square array is presented. The output signal of the antenna array is analyzed, showing quadrilinear model characteristics. The source parameters can be estimated from the matrices via low-rank decomposition based on uniqueness of PARAFAC decomposition. The algorithm first uses the quadrilinear alternating least square (QALS) algorithm to estimate the direction and frequency matrices. It then estimates frequencies according to Vandermonde characteristics of frequency matrix and least square principle. DOAs are estimated from the characteristics of direction matrix and the least square principle. The method does not need to search spectral
peaks or parameter pairs. Compared with ESPRIT and the trilinear model, the proposed algorithm has high precision in parameter estimation, and works well with small sizes. Simulation results verify its effectiveness.