Target Localization for MIMO Radar with Unknown Mutual Coupling Based on Sparse Representation

Abstract

Multiple-input multiple-output (MIMO) radars have attracted lots of attention for their special advantages. As a key issue, target localization method for MIMO radar has also been studied by lots of researchers. However, most of these methods are based on the ideal array model, and the practical unknown mutual coupling will make them have great performance degradation or even fail to work. Sparse representation (SR) has obtained a rapid development in the field of array signal processing over the past few years for its loose requirement for data amount and high-resolution feature. But only one-dimensional angle is considered, and the unknown mutual coupling will make the dictionary hard to construct when apply it for MIMO radar target localization. Even if the mutual coupling is compensated, the two-dimensional angle in MIMO radar will make the dictionary very huge, as well as enhance the inter-correlation of the dictionary. As a result, it brings higher complexity without performance improvement. However, according to our study, both the influences of mutual coupling and two-dimensional angle are occurred in space domain, but the output data of MIMO radar contains both the information in space and time domains. Thus, the influenced space domain is avoided and Doppler frequency in the time domain is utilized in this paper and not only the mutual coupling is avoided, but also a one-dimensional dictionary for the Doppler frequency can be obtained. The Doppler frequency can be firstly obtained via the sparse recovery technique, and meanwhile, the influenced direction matrix can be acquired from the non-zero rows of the recovered matrix. According to the special characteristic of the mutual coupling in uniform linear array (ULA), select the special rows of the direction matrix to eliminate the effect of mutual coupling and estimate the angles. The estimated target parameters (two-dimensional angles and Doppler frequency) are automatically paired, and the proposed algorithm has better estimation performance and can detect more targets than the subspace-based method and SR-based method. Simulation results verify the effectiveness of the algorithm.