Human gesture imitation using the whole body of a Nao robot and classification with neural networks

Abstract

This thesis completed an application for controlling a Nao robot to imitate human gestures and classification using the Kinect sensor. The main goal of the task was to manipulate the whole body of the robot while also keeping the balance and support changes needed. A project work was only about control Nao with only four upper joints of the body part and therefore there was no balancing problem. The Nao robot can become free to use both of its legs and change the support due to human movement. The second section of the thesis implemented a convolutional neural network to build a system for training and testing human daily activities to observe the advantages of joint data collected by Kinect sensor.