A novel method employing an ESPRIT and Kalman filter is proposed for the one-dimensional direction finding problem of MIMO radar. The rotational invariance among the multiple equivalent virtual subarrays of MIMO radar is exploited to derive multiple estimates of the direction of each target through ESPRIT. The multiple estimates of each target are fused through a Kalman filter to refine the accuracy of estimation. Numerical examples demonstrate the effectiveness of the proposed method.Introduction: By transmitting multiple orthogonal signals via multiple sensors and separating their echoes through matched filter banks, MIMO radar can extend the array aperture by virtual sensors [1], from which the accuracy of direction finding would benefit. The computationally efficient ESPRIT algorithm has been applied to the one-dimensional direction finding problem of bistatic MIMO radar in [2] and the reference therein, which is also applicable to the one-dimensional direction finding problem of monostatic MIMO radar concerned in this Letter. However, the estimation and eigen-decomposition of the covariance matrix of the huge virtual sensor array as required by the method in [2] are still computational complex. In this Letter, the virtual sensor array of monostatic MIMO radar is decomposed into multiple identical virtual subarrays [3]. The rotational invariance among these virtual subarrays is exploited to derive multiple estimates of the one-dimensional direction of each target through a modified ESPRIT algorithm. Then, the multiple estimates are fused through a Kalman filter to refine the estimation accuracy. Moreover, the sample multiplexing based on the symmetric data structure of MIMO radar further improves the estimation accuracy. The proposed method is more computationally efficient and accurate than the method presented in [2].

A novel signal subspace reconstruction method for multiple-input multiple-output (MIMO) radar is proposed. The developed method realises signal subspace resconstruction by the eigendecomposition of two low dimension matrixes. It provides lower computational complexity and fewer training samples compared to the full DOF method. Simulation results are given to demonstrate the effectiveness of the method.Introduction: Multiple-input multiple-output (MIMO) radar systems can transmit orthogonal waveforms to enhance resolution [1] and parameter identifiability [2] and detection performance [3] by greatly increasing the degrees of freedom (DOF) of the system. It was shown in [4] that paramaters estimation with the full DOF of the MIMO radar can achieve the highest resolution capability. However, the DOF of MIMO radar is often proportional to the product of M (the number of transmitters) and N (the number of receivers) and it can be very large. Thus the estimation of the covariance matrix and its eigendecomposition require a great number of training samples and very high computational cost. In this Letter, we propose a signal subspace reconstruction method for MIMO radar which can significantly reduce computational cost and the number of training samples.

No abstract