Automatic Engagement Detection for Social Situation Assessment
Title: Automatic Engagement Detection for Social Situation Assessment
Authors: Nicola Webb (University of the West of England); Manuel Giuliani (University of the West of England); Severin Lemaignan (PAL Robotics);
Year: 2022
Citation: Webb, N., Giuliani, M., Lemaignan, S., (2022). Automatic Engagement Detection for Social Situation Assessment. UKRAS22 Conference “Robotics for Unconstrained Environments” Proceedings, 58-59. doi: 10.31256/Mm8Xb7O
Abstract:
Abstract—In previous work, we collected simple interaction data through an online game. This collection method meant there was limited signal collection. From this, we created a metric to measure visual social engagement. In this work, we wish to collect an in-person dataset where we can capture a greater number of social signals. With this, we look to enrich our metric to investigate whether we can create more accurate
profiles of engagement. Also this dataset looks to capture dynamic group interactions, both human-human and human-robot, with participants being asked to complete a series of tasks both individually and alone. From this data, we will create a model of social engagement to automatically detect engagement levels, in order to improve a robots social awareness.