Commanding a drone through body poses, improving the user experience

In this work, we propose the use of Multimodal Human-Computer Interfaces (MHCI) through body poses to command a drone in an easy and intuitive way. First, the human-user pose is recovered from a video stream, with the help of the open source library Open Pose. Then, a Support Vector Classifier (SVC), trained to distinguish between different body poses, is used to interpret eleven different human poses as the most important high level commands to the drone. The proposed strategy was successfully implemented to remotely control a drone through a web interface, where the user can interact with a drone in a remote location using only a web interface. Real-time experiments were carried out with fourteen different volunteers, selected to represent different segments of the population, with different ages, gender, experience with technology, socioeconomic class, etc., in order to evaluate the user experience with the help of a User Experience Questionnaire (UEQ), demonstrating satisfactory results. The study suggests that the use of the proposed MHCI received good acceptance between the participants, even in users without previous experience with drones, and received excellent scores in terms of attractiveness, stimulation and novelty from most of the volunteers.