Error Correction Output Code-Based Radar Platform Motion Type Classification

Ozmen E., Cogun F., Alp Y. K., Altiparmak F.

IEEE Radar Conference (RadarConf), ELECTR NETWORK, 8 - 14 May 2021 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1109/radarconf2147009.2021.9455259
  • Country: ELECTR NETWORK
  • Keywords: Electronic Warfare, Multitask Learning, Support Vector Machine, Machine Learning, Error-Correcting Output Codes, Multiclass Classification
  • Hacettepe University Affiliated: Yes

Abstract

In this paper, radar parameters measured by EW systems are aimed to determine the platform types with which they are related. First, an efficient preprocessing method is applied, which involves quantifying range values and grouping class values to improve classification performance. Then, Multi-task Learning (MTL) neural network, Support Vector Machine (SVM) and Error-Correcting Output Codes(ECOC) techniques are applied to the problem. It is observed that ECOC technique outperforms SVM and MTL technique. It is considered that the result of the classification process can be used in the selection of the type and parameters of the filter used for tracking.