DOA esitmation based on support vector machine — Robustness analysis on array errors

Support vector machine(SVM) has gained good performance in classification. We treat the DOA estimation problem as a multi-class classification problem, and solve it by SVM. Train samples generated from array output data with known directions are used to train the SVM and construct classifiers, and then the classifiers will evaluate the test sample generated from unknown direction and derive the final DOA estimation result. The robustness for array errors is analyzed for the DOA estimation based on SVM. Simulation results are presented to confirm the robustness of the algorithm.