Please use this identifier to cite or link to this item: http://hdl.handle.net/10609/146088
Title: Disseny i implementació d'un sistema d'interfície natural d'usuari basat en càmera de profunditat i aprenentatge profund
Author: Ortiz Castelló, Vicent
Tutor: Vilaplana Besler, Veronica
Others: Morán Moreno, Jose Antonio  
Abstract: Different intelligent user interfaces have become widespread in our daily lives, including those based on touch and voice systems because they are versatile and non-invasive. For example, smartphones and virtual assistants are now everywhere. However, few current solutions interpret body gestures because high-performance systems are often expensive, and the cheaper ones, which work with visible spectrum sensors, depend heavily on the variability of the capture conditions to which they are subjected, especially to lighting and working distance, and raise privacy concerns. This project designs and implements a low-cost natural user interface based on Intel RealSense depth camera and deep learning models running on an NVIDIA Jetson Nano, a low-consumption mobile computing platform. Model training is modular and may be done on the same mobile platform or on more powerful hardware to save time. The solution allows robust human-machine interaction at a distance, in any light condition -even in the dark- and the set of detected poses can be customised by taking a video sequence of each new gesture and retraining the model. Possible applications include interacting with different types of smart devices -computers, televisions, heating elements, and home automation, among others- with a high level of privacy since no visible spectrum information is recorded.
Keywords: natural user interface
deep learning
depth camera
Document type: info:eu-repo/semantics/bachelorThesis
Issue Date: Jun-2022
Publication license: http://creativecommons.org/licenses/by-nc-nd/3.0/es/  
Appears in Collections:Bachelor thesis, research projects, etc.

Files in This Item:
File Description SizeFormat 
viorcasTFG0622memòria.pdfMemòria del TFG4,52 MBAdobe PDFThumbnail
View/Open