Konferenzbeitrag
GAIT3: An Event-based, RGB and Thermal Gait Database
Vorschaubild nicht verfügbar
Volltext URI
Dokumententyp
Text/Conference Paper
Zusatzinformation
Datum
2022
Autor:innen
Zeitschriftentitel
ISSN der Zeitschrift
Bandtitel
Quelle
Verlag
Gesellschaft für Informatik e.V.
Zusammenfassung
Identifying people by their gait has gained popularity in the last twenty years. Recent gait
recognition methods use acquisitions extracted from advanced sensors such as cameras, depth sensors,
microphones, etc. Recently, event-based cameras, a new family of cameras, are gaining popularity. They are
vision sensors that differ completely from conventional cameras: instead of capturing images at a fixed rate,
they asynchronously measure per-pixel brightness changes generated by moving objects. This motivated
us to use it for individual recognition by gait.
In this paper, we provide means for multimodal gait recognition, by introducing the “Event-based, RGB,
and Thermal Gait” database. This database is the first that contains event-camera acquisition, simultaneously
with conventional RGB and thermal videos. It contains recordings of people in three variations: normal
walking, quick walking, and walking with a backpack.
We also present experiments using a baseline algorithm based on gait energy images adapted to event-based
camera output. Then we present a comparative experiment against RGB and thermal videos, using the
same algorithm, that shows an advantage for event-based data.