The Central DOA Estimation Algorithm Based on Support Vector Regression for Coherently Distributed Source

摘要

In this paper, the problem of estimating the central direction of arrival (DOA) of coherently distributed source impinging upon a uniform linear array is considered. An efficient method based on the support vector regression is proposed. After a training phase in which several known input/output mapping are used to determine the parameters of the support vector machines, among the outputs of the array and the central DOA of unknown plane waves is approximated by means of a family of support vector machines. So they perform well in response to input signals that have not been initially included in the training set. Furthermore, particle swarm optimization (PSO) algorithm is expressed for determination of the support vector machine parameters, which is very crucial for its learning results and generalization ability. Several numeral results are provided for the validation of the proposed approach.