Multidimensional multirate DOA estimation in beamspace

This paper presents a beamspace version of ESPRIT for uniform rectangular arrays that supports closed-form 2-D angle estimation, automatically couples the two components of the source directions, and works with any front end beamformer. The proposed algorithm is based on the observation that beamspace noise eigenvectors can be transformed to vectors that are bandpass and have spectral nulls at the inband source locations. This facilitates multirate processing (involving modulation to baseband, filtering, and decimation) and yields a space with dimensionality equal to the number of beams used to probe the subband rather than the number of elements in the sensor array. The MUSIC algorithm can be applied to this noise subspace. Alternatively, a transformation matrix can be computed a priori, which maps the beamspace signal eigenvectors to the corresponding signal subspace that has the ESPRIT structure. The TLS-ESPRIT algorithm is then modified to obtain the two directions for each source from a single eigenvalue-eigenvector pair. Hence, they are automatically coupled