Research on Support Vector Machines Framework for Uniform Arrays Beamforming

In order to explore a new optimization method for array signal beamforming, after studying the mathematical principal of the Support Vector Machine (SVM) algorithm and its primal cost function, we apply the modified cost function to the uniform array beamforming and minimize the constrained items by means of the method of Lagrange multipliers. Ultimately, the new SVM-based optimizing beamforming approach is proposed, which is used to optimize the corresponding parameters of array beamforming. The Support Vector Machine framework for uniform arrays beamforming is then established. Simulation results show that the SVM-based optimizing beamforming approach can not only approximate conventional one well in the case of free noise condition and small data sets, but also improve the generalization ability and reduce the computation burden. Also the sidelobe level of both linear and circular arrays by the SVM algorithm is improved sharply than the conventional one. The SVM-based beamforming is superior to the conventional one no matter linear and circular arrays, single and two non-coherent sources. The good performance shown on these different scenarios suggests that other beamforming optimization problems can be stated from this SVM framework. Compared with the conventional beamforming approaches, it provides a new and effective technique for the optimization design of beamformer.