A Multi-modal Audience Engagement Measurement System
Authors: Miguel Sanz Narrillos Stefano Masneri
Date: 14.03.2021
Abstract
During live events the organizers often would want to deploy audience engagement systems to analyze people behaviour, perform user profiling or modify the show according to the participant feedback. Such systems usually need computer vision algorithms whose performance are severely affected by constraints such as illumination and cameras position. In this paper we present a fully automatic audience engagement system, optimized for live music events with rapidly changing illumination conditions. The system uses a multi-modal approach which combines wireless-based person detection together with computer vision algorithms for pose and face analysis. We show that such hybrid approach, while running in real-time, performs better than standard approaches that only employ computer vision techniques. The system has been tested both in a laboratory environment as well as in a concert hall and it will be deployed in distributed live events.
BIB_text
title = {A Multi-modal Audience Engagement Measurement System},
pages = {354-377},
keywds = {
Multi-modal analysis Wireless detection Computer vision Audience engagement
}
abstract = {
During live events the organizers often would want to deploy audience engagement systems to analyze people behaviour, perform user profiling or modify the show according to the participant feedback. Such systems usually need computer vision algorithms whose performance are severely affected by constraints such as illumination and cameras position. In this paper we present a fully automatic audience engagement system, optimized for live music events with rapidly changing illumination conditions. The system uses a multi-modal approach which combines wireless-based person detection together with computer vision algorithms for pose and face analysis. We show that such hybrid approach, while running in real-time, performs better than standard approaches that only employ computer vision techniques. The system has been tested both in a laboratory environment as well as in a concert hall and it will be deployed in distributed live events.
}
isbn = {978-303071157-3},
date = {2021-03-14},
}
