Hybrid Sparse Array Beamforming Design for General Rank Signal Models

Abstract

The paper considers sparse array design for receive beamforming achieving maximum signal-to-interference plus noise ratio (MaxSINR) for both single point source and multiple point sources, operating in an interference active environment. Unlike existing sparse design methods which either deal with structured environment-independent or non-structured environment-dependent arrays, our method is a hybrid approach and seeks a full augumentable array that optimizes beamformer performance. This approach proves important for limited aperture that constrains the number of possible uniform grid points for sensor placements. The problem is formulated as quadratically constraint quadratic program (QCQP), with the cost function penalized with weighted l1-norm squared of the beamformer weight vector. Simulation results are presented to show the effectiveness of the proposed algorithms for array configurability in the case of both single and general rank signal correlation matrices. Performance comparisons among the proposed sparse array, the commonly used uniform arrays, arrays obtained by other design methods, and arrays designed without the augmentability constraint are provided.

Learning Sparse Array Capon Beamformer Design Using Deep Learning Approach

“…approach because both incorporate the apriori knowledge of interference parameters. Therefore, in comparing the data dependent designs, it is found that SDR design (also the SDR-symmetric [20]) is comparable to the DNN-EN design, with the DNN-SBSA is marginally suboptimal with the average performance degradation of 0.37 dB. This slight performance trade off is readily leveraged by the real time realization of the DNN-SBSA algorithm implementing the Capon beamformer in time frames of the order of few milli-seconds.…”

“…The performance of the proposed SBSA, DNN-EN and DNN-SBSA are compared with existing work on sparse array design which is based on SDR and SCA approaches [19], [20]. It is clear from Fig.…”

“…Here, ||.|| 0 determines the cardinality of the weight vector w. This is a combinatorial optimization problem and can be solved by enumerating over all possible sensor locations. Several different approaches have been developed to mitigate the computational expense of the combinatorial search by either exploiting prior given information on the interference parameters, such as respective DOAs, or employing datadependent algorithms realized through the SDR and SCA algorithms [15], [20]. These algorithms, however, have high Fig.…”

“…In both cases, several iterative algorithms have been proposed to optimize the sparse array beamformer design. Although, convex based optimization algorithms, such as semidefinite relaxation (SDR) and successive convex approximation (SCA) have been developed to yield sparse configurations with desirable beamforming performances [19], [20], real time implementations of these algorithms remain limited due to the relatively high computation cost. The problem becomes more pronounced in rapidly changing environments which result from temporal and spatial non-stationary behaviors of the sources in the field of view.…”

Vector Hydrophone Array Design Based on Off-Grid Compressed Sensing

“…From the perspective of array structure, these methods can be divided into two major categories. The first class is the sparse selection approach [1][2][3][4][5][6], which is based on arrays that are uniformly arranged (usually array elements spaced half a wavelength apart), and a portion of the array elements are selected to be unexcited based on the optimization results. The second class is the sparse array design method [7][8][9][10][11][12][13][14][15][16][17][18][19][20], in which the array elements are randomly arranged on the array plane.…”