Multiple main user recognition
Hye-Jin S. Kim, Il-Kwon Jeong · 2018 20th International Conference on Advanced Communication Technology (ICACT) · 2018
We present a method for selecting main users among multiple candidates. For human-machine interaction, a machine choose a single or multiple users to interact with them. The proposed method provides an ordered user probability that indicates whether one can be a main user. We estimate various features such as distance to a user, eye detection, gaze direction, moving velocity, and acceleration. Experiments show that our algorithm effectively differentiates main users from the others.