Unit:
Department of Informatics and TelecommunicationsΠληροφορική
Author:
Ioannides Yiannakis
Supervisors info:
Αλωνιστιώτη Αθανασία
Αναπληρώτρια Καθηγήτρια
Τμήμα Πληροφορικής και Τηλεπικοινωνιών
Εθνικό και Καποδιστριακό Πανεπιστήμιο Αθηνών
Original Title:
Ανάλυση σημάτων εγκεφαλογραφήματος για οδική ασφάλεια
Translated title:
Analysis of encephalogram signals for road safety
Summary:
in this work, the basic machine learning methodology is proposed to classify the condition
of the eyes (ie, eyes open or closed) using electroencephalographic data for road safety.
The idea is to compare and validate the basic Machine Learning (K-Nearest Neighbors
KNN) approach to the support vector machine (SVM) algorithm. EEG data were collected
using Emotiv Epoc+ headphones and each recording was manually labeled, containing
14 channels (recording columns) using a tag as open or closed eyes. The experimental
results confirm that the methodology of using the KNN provides a better prediction
accuracy of 73% using entropy as the extraction method. The proposed approach for the
equally important F1-Score gave the same results, using the KNN achieving 75%.
The structure of the work is as follows:
Chapter 1 highlights the importance of driver fatigue in road accidents and how the
detection of blindfolds can be critical. Then, in chapters 2 and 3, the network topology
and analysis of the entities of the experimental system for the exchange of data within it
(from the android device to the SQL database) are analyzed. Chapter 4 describes the
experiments performed on 71 students and the collection of data with the Emotiv Epoc +
tool. Chapter 5 lists the two feature extraction methods used, namely the fast Fourier
transform and the entropy. Chapter 6 studies the kNN and SVM algorithms from a
theoretical point of view (and less descriptively others such as the decision tree algorithm,
the random forest, etc.). Chapters 7 and 8 analyze the results of the experiments
compared to the metrics of accuracy, F1 score and Jaccard and confirm the superiority
of kNN over SVM.
Main subject category:
Technology - Computer science
Keywords:
neural networks, driving, eyes, encephalogram, machine learning, artificial intelligence
File:
File access is restricted only to the intranet of UoA.
final_thesis_Ioannides_Yiannakis.pdf
1 MB
File access is restricted only to the intranet of UoA.
