Classification of human activities on UWB radar using a support vector machine

Abstract

In this paper, we classify seven different human activities measured by a ultra wideband (UWB) radar using a Support Vector Machine (SVM). The classification is done using the time variation of a signature of a return from a human subject. This time varying signature is unique to a particular motion because human's returns vary based on the change in the orientation of their torso and limbs. We exploit this time variation of a human's radar signature in order to classify the human activity recorded by the radar. The signature is captured by the Principle Component Analysis (PCA). The Support Vector Machine (SVM) is proposed as a classifier. The training process and the resulting classification accuracy are reported.