DNN-based approach for identification of the level of attention of the TV-viewers using IoT network
Hamdi Amroun, M'hamed Hamy Temkit, Mehdi Ammi · 2017
In this paper, we define the concept of levels of attention towards a TV program. Then, we propose an approach to identify and classify them using a network of connected objects: a smartphone, a smart watch and a remote control. Seven participants were observed watching their favorite TV programs. Then each participant was asked to rate some of their own activities according to two situations: attentive to the TV program (Level 1) or not (Level 2). Attention levels were classified using a Deep Neural Network algorithm (DNN) after extracting a set of descriptors from the sensors' data of the three devices and fused them. The results show that the levels of attention were classified with an accuracy of 88.44% and 87.89% for the level 1 and level 2 respectively. This study shows that we can detect a level of attention towards a TV program with usual connected objects while remaining as intrusive as possible. This corresponds to real situations of sensing attention to a TV program.