Label Interpolation Based Deep Learning for Direction of Arrival Estimation

Shiwei Ren, Dingsu Xu, Weijiang Wang*

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

In this paper, an end-to-end framework is proposed for direction of arrival (DOA) estimation on the conditions of anechoic and low SNR environments via Deep Learning (DL). Spectrum labels generated by interpolation methods are used for regressors attainment in the training of neural networks. When the regressors converge properly, a peak finding strategy is proposed to estimate DOAs more precisely, which consists of the Gaussian smoothing, rough sampling and Least Squares method. Numerical experiments prove that this framework performs better than the existing DL methods.

Original languageEnglish
Title of host publicationProceedings - 2021 7th Annual International Conference on Network and Information Systems for Computers, ICNISC 2021
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages937-943
Number of pages7
ISBN (Electronic)9781665402323
DOIs
Publication statusPublished - 2021
Event7th Annual International Conference on Network and Information Systems for Computers, ICNISC 2021 - Virtual, Guiyang, China
Duration: 23 Jul 202125 Jul 2021

Publication series

NameProceedings - 2021 7th Annual International Conference on Network and Information Systems for Computers, ICNISC 2021

Conference

Conference7th Annual International Conference on Network and Information Systems for Computers, ICNISC 2021
Country/TerritoryChina
CityVirtual, Guiyang
Period23/07/2125/07/21

Keywords

  • DL
  • DOA estimation
  • spatial spectral peak search

Fingerprint

Dive into the research topics of 'Label Interpolation Based Deep Learning for Direction of Arrival Estimation'. Together they form a unique fingerprint.

Cite this