Characterization of postures to analyze people’s emotions using Kinect technology

Abstract

This article synthesizes the research undertaken into the use of classification techniques that characterize people's positions, the objective being to identify emotions (astonishment, anger, happiness and sadness). We used a three-phase exploratory research methodology, which resulted in technological appropriation and a model that classified people’s emotions (in standing position) using the Kinect Skeletal Tracking algorithm, which is a free software. We proposed a feature vector for pattern recognition using classification techniques such as SVM, KNN, and Bayesian Networks for 17,882 pieces of data that were obtained in a 14-person training sample. As a result, we found that that the KNN algorithm has a maximum effectiveness of 89.0466%, which surpasses the other selected algorithms.