Research On Robust Adaptive Beamforming Algorithms

Adaptive beamforming is one of the key technologies in array signal processing and has been widely used in sonar, communications, and medical imaging etc. The performance of the adaptive beamformer will be degraded by differences between the ideal signal environment and the real one such as random error, covariance matrix estimation error and steering vector error etc. Therefore, the research on the robustness of the adaptive beamformer is very essential.Diagonal loading is a common and effective method to improve the robustness of beamforming while its loading factor is not so easy to determine. Although researchers have presented many methods to calculate the diagonal loading factor for improving the performance of the beamformer, most of them are based on experience or need additional parameters. However, in most cases, these parameters are hard to be estimated. Since they play an important role in the beamforming algorithms, we prefer some non-parameter based schemes in practical applications.First, the mathematical model of the adaptive array and some measure methods are introduced in this paper, followed by some common beamforming algorithms and analysis of their advantages and disadvantages. Whereafter the theoretical analysis and simulations of the diagonal loading to verify the algorithms are presented and an improved automatic GLC-based robust algorithm is proposed. In the case of both steering vector errors and covariance matrix errors, the novel scheme can increase the SINR by adding a weight to adjust the magnitude of the loading factor. The weight is related to the steering vector errors. The simulation results show that the novel method achieves better performance without the knowledge of extra estimated parameters. Moreover, it can output higher SINR in the case of DoA mismatch.At last, some introduction of Dspbuilder is presented and FPGA behavioral level simulations of the key point of the algorithm are proposed. The simulation results show that the system architecture of the matrix calculation is correct.