Abstract: A nonlinear adaptive beamforming approach based on Least-Square Support Vector Regression (LS-SVR) is proposed to enhance the beamformers robustness against array model mismatch, constrained samples numerous interferences, etc. The approach has two highlights, one is a recursive regression procedure to compute the regression parameters on real-time, the other is a sparse mode based on novelty criterion, which can significantly reduce the size of the input samples. Applying the sparse model to LS-SVR beamforming leads to reduced computation complexity and better generalization capacity. The theory analysis and experimental results show that the proposed beamforming approach could improve array performance significantly over several classical linear beamforming methods.