| In the conventional signal processing approachs,the signal had be modeled as being stationary, but in real world applications, there are signals that are nonstationary,considering a special nonstationary signals – cyclo-stationary signals, this paper researches the “virtual” array structure of cyclo-stationary signals and the problem of direction-of-arrive(DOA) estimation under the circumstances that more source exist.The commonly used DOA estimation approach based on the array expansion is the higher order cumulants algorithm, the increased order implies larger data lengths for reliable estimation of statistics, and the computational complexity is greater,Therefore, this paper intend to reaserch a DOA estimation approach based on the array expansion and lower computational complexity under the guidance of current theory, which is about a study based on the virtual array structure and its DOA estimation of cyclo-stationary signals via Khatri-Rao subspace. To avoiding high order cumulants, the virtual array structure is based on the feature of Khatri-Rao product provided inherently in the second-order statistics model, receiving the full development of cyclo-stationary signals subspace feature provided in the second-order statistics model, afterward, this paper introduce a dimension reduction approach for reduce the problem dimension under the premise of effective degrees of freedom. The higher order cumulants algorithm is compared further.Theoretical studies revealed that the proposed Khatri-Rao subspace approach can be applied to the cases that sensors is less than sources to achieve a virtual extension,while the computational efficiency will not be substantially increased; Simulation shows that the KR subspace approach can estimate DOA under the circumstances that more source exist, compared to the method based on higher order cumulants(HOCS),the running time of this algorithm is less, but the rejection of colored noise is not strong. |
PDF Full Text Request