Extraction of information from videos, application to human behavior understanding
We
study the feasibility of capturing and analyzing human behaviors (emotions, abnormal events, suspect
events, flow, eye gaze) from video and
complex data, considering the context and profiling capture.
The objective is to understand and predict human behaviors in
real-life and controlled environments (e.g., personal environments,
airports, malls, etc.). The inference of prediction (human intentions)
could be used to assist a human in his/her day-to-day tasks by adapting
the system’s response to his/her needs.
The study is
validated in several application domains related to security (e.g.,
detecting collapsed situations in escalator exits of an airport, panic,
potential abnormal events), and marketing (e.g., estimating the people
flow in front of a shop, hot areas inside the shop).