Continuous Human Activity Classification with Radar Point Clouds and Point Transformer Networks

Conference paper (2023)

Authors

N.C. Kruse

F. Fioranelli

Alexander Yarovoy

DOI: https://doi.org/10.23919/EuRAD58043.2023.10289503

Machine learning Radar Human activity recognition Point cloud processing

More Info

expand_more

To reference this document use:

http://resolver.tudelft.nl/uuid:25bbc0d7-038e-44d7-88f8-d4c64f4f034a

Published Date

2023

Language

English

Reuse Rights

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.

Abstract

Due to numerous benefits, radar is considered as an important sensor for human activity classification. The problem of classifying continuous sequences of activities of unconstrained duration has been studied in this work. To tackle this challenge, a radar data processing method utilizing point transformer networks has been proposed. The method has been experimentally verified on a dataset of human activities, and experiments have been performed to determine its optimal implementation. Promising preliminary results on a 9-class dataset show test accuracy and macro F-1 scores in the range of 83% and 73% respectively.

Files

Continuous_Human_Activity_Clas... (.pdf)

(.pdf | 1.37 Mb)

- Embargo expired in 26-04-2024