Facial expressions recognition with CNN and its application in an assistant humanoid robot

Abstract

In this paper, the development of a Convolutional Neuronal Network (CNN) of three layers for the recognition of six different facial expressions and its implementation in an assistant humanoid robot is presented. A database is generated with different facial expressions, supervised training is used and accuracy of the Convolutional Neuronal Network employed is determined. The Convolutional Neural Network implementation in the robot was possible with a Raspberry Pi 3 model B which also aides in the execution of different movements depending on the facial expression that is recognized.