High-Dimensional Range Profile Geometrical Visualization and Performance Estimation of Radar Target Classification via a Gaussian Mixture Model

Thomas Boulay (1, 2) , Ali Mohammad-Djafari (2) , Nicolas Gac (2) , Julien Lagoutte (1)

1 Thales Research and Technology [Palaiseau]
2 L2S - Laboratoire des signaux et systèmes

In this paper, a method of data visualization and classification performance estimation applied to target classification is proposed. The objective of this paper is to propose a mathematical tool for data characterization. The principle is to use a non linear dimensionality reduction technique to describe our data in a low-dimensional space and to model embedding data by Gaussian mixture model (GMM) to estimate classification performance graphically and analytically.

Domaines: Sciences de l'ingénieur [physics]

Thomas Boulay, Ali Mohammad-Djafari, Nicolas Gac, Julien Lagoutte. High-Dimensional Range Profile Geometrical Visualization and Performance Estimation of Radar Target Classification via a Gaussian Mixture Model. Geometric Science of Information, Aug 2013, Paris, France. pp.829-836, ⟨10.1007/978-3-642-40020-9_93⟩. ⟨hal-00931023⟩