RSS-Based improved DOA estimation using SVM

Abstract

Many applications like wireless communications, radars, objects detection need precise localization particularly in 5G communications. An approach to improve direction of arrival (DOA) estimation based on machine learning, correlation matrix and received signal strength (RSS) is proposed. The proposed method relies on a fine feature selection to rise generalization capability of a support vector machine (SVM) and received signal strength (RSS) to further enhance the generalization capability and angle estimation precision. While standard usage of SVM network with correlation matrix leads to 50% generalization capability the proposed approach rises the performances up to 98%. The approach is tested with success for the estimation of a two-dimensional DOA.