(A) mobile system design study of monitoring child enticement

Abstract

Keeping child safe is a critical mission for parents and also society. Despite the importance, there are several threats that can make child dangerous. We point out that enticing child can be a starting mark of those dangers, in particular, child abduction and sexual assaults. In this work, as far as we understand, we firstly propose a monitoring system that enables to catch child enticing situation. For a child, enticing is difficult to fully understand since the child usually has no ability to judge a situation exactly. We observe several experimental videos that reveal children are vulnerable for determining whether a temptation (e.g. "Do you want a candy?") is intentional or not. Conducting interviews with parents who use kids wearable devices, we understand limitations of current devices. Implementing a monitoring system with a commercial smartwatch, we make three scenarios including enticing, street talking, and visiting a store. With modeled scenario, we present that the proposed design enables to capture enticing scenario. We exploit deep learning for voice separation and the Android APIs for walking detection to judge the child enticement scenario.