Coherent SVR Learning for Wideband Direction-of-Arrival Estimation
Liuli Wu, Zhitao Huang
Abstract
Liuli Wu, Zhitao Huang
Abstract
In this letter, we propose a coherent support vector regression (SVR) scheme to address the wideband direction of arrival (DOA) estimation problem. This learning-based method deals with wideband DOA estimation by treating it as a function approximation issue. The proposed approach first decomposes the wideband array outputs into several narrowband components, then approximates the functional relationship between the decomposed narrowband data and the DOA with coherent SVR scheme through training. The trained function is then capable of estimating the DOA when wideband signal with unknown impinging direction arrives. We prove the effectiveness and superiority of the presented method by simulation experiments. Simulation results show that the new technique has a better performance in terms of estimation errors than the conventional broadband DOA estimation method, especially in demanding scenarios with low SNR and limited snapshots. Moreover, the proposed approach also relaxes the unambiguous array element-spacing restrictions, i.e., it has extended the frequency range of wideband signals where direction finding without angle ambiguity is achievable.
OpenAlex reports 81 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
In this letter, we propose a coherent support vector regression (SVR) scheme to address the wideband direction of arrival (DOA) estimation problem. This learning-based method deals with wideband DOA estimation by treating it as a function approximation issue. The proposed approach first decomposes the wideband array outputs into several narrowband components, then approximates the functional relationship between the decomposed narrowband data and the DOA with coherent SVR scheme through training. The trained function is then capable of estimating the DOA when wideband signal with unknown impinging direction arrives. We prove the effectiveness and superiority of the presented method by simulation experiments. Simulation results show that the new technique has a better performance in terms of estimation errors than the conventional broadband DOA estimation method, especially in demanding scenarios with low SNR and limited snapshots. Moreover, the proposed approach also relaxes the unambiguous array element-spacing restrictions, i.e., it has extended the frequency range of wideband signals where direction finding without angle ambiguity is achievable.
Key concepts: Wideband, Direction of arrival, Narrowband, Computer science, Broadband, Algorithm, Direction finding, Function (biology)