Capturing, Modelling, Analyzing and providing Feedback in Martial Arts with Artificial Intelligence to support Psychomotor Learning Activities

Casas Ortiz, Alberto. (2020). Capturing, Modelling, Analyzing and providing Feedback in Martial Arts with Artificial Intelligence to support Psychomotor Learning Activities Master Thesis, Universidad Nacional de Educación a Distancia (España). Escuela Técnica Superior de Ingeniería Informática. Departamento de Inteligencia Artificial.

Ficheros (Some files may be inaccessible until you login with your e-spacio credentials)
Nombre Descripción Tipo MIME Size
Casas_Ortiz_Alberto_TFM.pdf Casas_Ortiz_Alberto_TFM.pdf application/pdf 3.73MB

Título Capturing, Modelling, Analyzing and providing Feedback in Martial Arts with Artificial Intelligence to support Psychomotor Learning Activities
Autor(es) Casas Ortiz, Alberto
Abstract This Master’s Thesis explores how Artificial Intelligence (IA) can assist in the learning of psychomotor activities, specifically the learning of a martial art, through the development of an AI-based application that executes in an Android device. Martial arts are an interesting domain for this because it encompasses most of the characteristics that can be found in other psychomotor activities. Different methods for capturing, modelling and analyzing human motion, as well as providing feedback to the user have been reviewed. In addition, another bibliographical review of 27 publications has been carried out to evaluate till what extend these methods have been already applied to martial arts. For this research work, inertial methods have been selected for capturing motion. In particular, the inertial sensors of an Android device have been used for capturing the execution of a set of movements of American Kenpo Karate from 20 volunteers. The captured data was then modeled, by segmenting and labelling the movements, and smoothing the time series using Exponentially Weighted Moving Averages. The resulting dataset, formed by 240 movements, was then used for training and comparing three neural network-based classifiers: FC-ANN, 1D-CNN and LSTM. Neural networks were selected because of their ability of learn complex functions and the fact that some neural network architectures have been created specifically for analyze time series. Further, the weights learned by a neural network can be transferred to other domains through the technique known as transfer learning. Obtained results suggest that LSTM is the type of neural network that can better classify the movements studied, obtaining an accuracy of 1.0 in the training set, and an accuracy of 0.94 in the testing set. For demonstrating that those methods can be applied, an AI-based real-time Android application has been developed. This application employs the studied methods, as well as a feedback strategy created using the results of a questionnaire carried out with the purpose of identifying the issues that online learning of a psychomotor activity can entails. The application has then been tested, generating a good impression in the users. Following the open science philosophy, all contributions are shared in the GitHub repository.
Notas adicionales Trabajo de Fin de Máster. Máster Universitario en I.A. Avanzada: Fundamentos, Métodos y Aplicaciones. UNED
Materia(s) Ingeniería Informática
Palabra clave martial arts
human motion capture
human motion modeling
human motion analysis
sequence analysis
feedback strategies
american kenpo karate
Editor(es) Universidad Nacional de Educación a Distancia (España). Escuela Técnica Superior de Ingeniería Informática. Departamento de Inteligencia Artificial.
Director/Tutor Santos, Olga C.
Fecha 2020-06-01
Formato application/pdf
Identificador bibliuned:master-ETSInformatica-IAA-Acasas
http://e-spacio.uned.es/fez/view/bibliuned:master-ETSInformatica-IAA-Acasas
Idioma eng
Versión de la publicación acceptedVersion
Nivel de acceso y licencia http://creativecommons.org/licenses/by-nc-nd/4.0
info:eu-repo/semantics/openAccess
Tipo de recurso master Thesis
Tipo de acceso Acceso abierto

 
Versiones
Versión Tipo de filtro
Contador de citas: Google Scholar Search Google Scholar
Estadísticas de acceso: 119 Visitas, 59 Descargas  -  Estadísticas en detalle
Creado: Mon, 20 Sep 2021, 20:49:28 CET