Home // International Journal On Advances in Life Sciences, volume 10, numbers 3 and 4, 2018 // View article
Authors:
Bruno Barbosa
Antonio Neves
Sandra C. Soares
Isabel Dimas
Keywords: Pose Estimation; Digital Image; Emotions; Skeleton Detection
Abstract:
Recent advances in the study of emotions show that body expressions or body postures are also a relevant non-verbal cue for communicating emotions, with recognition rates similar to those transmitted by facial expressions. Despite the tremendous technological advances in the last decades, there is still room for improvement in the definition of how posture features can be related to different emotions, using technological solutions. This is a complex task due to the fact that it is difficult to associate an emotion with a certain posture and there are a large variety of emotions, such as sadness, joy, anger, fear, surprise, disgust, among others. In this paper, we present a computer vision system that is able to extract characteristics regarding the human body and generate data that allow the evaluation of a given posture. We propose the use of a recently developed algorithm to extract the body posture based on digital color images, which has presented optimal results in several domains, for the recognition of emotions from human body postures. Based on this algorithm, a system was developed that allows, in a non-invasive way, to extract characteristics of the human posture in order to perceive the emotions associated with each posture. We present experimental results in a real scenario, showing promising results regarding the extraction of characteristics obtained from the human body (displacements, areas, and overlaps), confirming the effectiveness of the developed system.
Pages: 198 to 210
Copyright: Copyright (c) to authors, 2018. Used with permission.
Publication date: December 30, 2018
Published in: journal
ISSN: 1942-2660