A novel method for two-dimensional sinusoidal signal retrieval

For a two-dimensional (2-D) harmonic signal that consists of L plane waves, we verified that first eigenvector corresponding to the maximum eigenvalue of the autocorrelation matrix that belongs to span of the signal vector. Arranging the first eigenvector, we gain a matrix instead of the data matrix in MEMP method. Then we construct an enhanced matrix from above matrix. It was found that there are signal subspace and noise subspace in the singular value decomposition of the enhanced matrix. We present a spectral estimation method based on the noise subspace of enhanced matrix. The method possesses high statistical stability and high resolution. The effectiveness of our methods is demonstrated through some simulation examples.