Gesture Recognition in an IoT environment: a Machine Learning-based Prototype

The spread of IoT and wearable devices is bringing out gesture interfaces as a solution for a more natural and immediate human-machine interaction. The "Seamless" project is an industrial research and development experience, aimed to build a virtual environment where data collected by IoT sensors can be navigated through a gesture interface and virtual reality tools. This paper presents the portion of the project concerning gesture analysis, focused on the problem of automatically understanding a set of hand gestures, in order to give commands through a wearable control device. The tackled issue is to build a real time gesture recognition system based on inertial data, that can easily adapt to different users and to an extensible set of gestures. This gesture data variability is addressed by means of a supervised Machine Learning approach, that allows adapting the system response to different gestures and to various ways of performing them by different people. A context-aware adapter allows interfacing the gesture recognition system to various applications.