Recognition of Human Activity Using Paired Connected Objects
Hamdi Amroun, MHamed Hamy Temkit, Mehdi Ammi · 2017
This paper proposes a method for recognizing human activity using paired connected objects: a smartphone paired with a connected remote control. The approach consists of classifying two types of activities: Making a call phone (Call) and managing TV with the paired smartphone (manage TV). Seven participants wore the smartphone, once paired with the connected remote control, while they sit in front of the TV. A classification of these two activities was made by a Deep Neural Network algorithm (DNN), without data preprocessing. Results show that a classification accuracy of 99.63% has been achieved. Our method can be used to help identify the owner of a paired smartphone with a remote control to protect connected remote control data from any act of mailvailance, such as subscriptions to paid movies or TV channels.