Publication: Development of a bodyguard application in Android
Loading...
Identifiers
Publication date
2018-07-03
Defense date
2018-07-03
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
In this thesis it is going to be explained the process to develop a bodyguard application
using Android. This application should be able of properly detecting whether the user is
suffering a physical aggression or not and, if so, send an email alert to an emergency contact
previously established. It will be also tried to extend the classification to additional
movement patterns.
For that, a first process involving the collection of the training samples will be performed.
Once a proper dataset is obtained, different classifiers such as kNN, Multilayer
Perceptron or Support Vector Machines will be explored to decide which one fits better
to the activity recognition and, after that, a feature reduction process is applied to reduce
the number of attributes in the classification with the purpose of making the application
more efficient with respect to the number of calculations. Finally, an optimization of the
selected classifiers comes to obtain the highest accuracy possible.
In the end, it is obtained an Android application that properly differentiates between
aggression and non-aggression movements and capable of sending the email alert when
needed in a low amount of time.With respect to the classification of additional movement
patterns, it was not achieve a high precission for all of the classifiers considered due to
inaccuracy problems in the training data that will be explored in future works.
Description
Keywords
Development of applications, Android, Weka, Supervised learning, Multilayer Perceptron, kNN, Activity recognition