CJ: An intelligent robotic head based on deep learning for HRI

Abstract

The aim of this article is the design and construction of a robotic head intended for classification and recognition applications using algorithms based on Deep Learning. The robotic head works by performing a classification on a number of 1000 different objects thanks to the use of a convolutional neural network (CNN) trained with ImageNet. On the other hand, applying the transfer learning technique in a CNN, allows the recognition of faces on a database of 185 people. The robotic structure also has a voice recognition system, which allows the user to interact with the robot using different voice commands. The construction of the structure is mainly divided into 3D printed components, servomotors, and an Arduino UNO microcontroller, which oversees all the movement that the structure performs.