Multidimensional Array Interpolation Applied to Direction of Arrival Estimation

In MIMO communications systems, the data has a natural multidimensional structure composed of time, frequency and space dimensions. Recently, multidimensional techniques that take into account the data multidimensional structure have been proposed for model order selection, parameter estimation and prewhitening. These multidimensional techniques require an array with a PARAFAC structure. However, in practice, building antenna arrays with precise geometries is not feasible. In this paper, we propose a multidimensional array interpolation scheme that forces a real imperfect array to become a PARAFAC array. Once the multidimensional interpolation is successfully performed, advantages such as increased identifiability, separation without imposing additional constraints and improved accuracy can be exploited. Numerical simulations show that the proposed method provides improved DOA estimation accuracy when a PARAFAC technique is applied to an originally nonPARAFAC array.