Document Details

Document Type : Thesis 
Document Title :
TRANSFER LEARNING ALGORITHM FOR LOW-COUNT TRAINING TRIALS WITH DEEP LEARNING IN ELECTROENCEPHALOGRAPHY
خوارزمية نقل التعلم لتجارب التدريب المنخفضة مع التعلم العميق على الانفعال الكهربائي للدماغ
 
Subject : faculty of Engineering 
Document Language : Arabic 
Abstract : The field of Artificial intelligence is rapidly evolving in daily basis and the fact of that brain computer interface (BCI) enables the computer to imitates the human brain processes. There are multiple types of algorithms but in this thesis will be concern about Neural Network and Transfer Learning. A new transfer learning methodology were proposed in the field of Electroencephalography (EEG) signal processing in order to study the performance. This thesis proposed approaches to transfer learning from four labels in which the source class is different from the target class in terms of valence, arousal, dominance and liking labels. This is achieved by two approaches, using Discrete Wavelet Transform (DWT) and the other way using Power Spectral Density (PSD) as feature extraction for EEG signals from Database for Emotion Analysis using Physiological Signals (DEAP). Transfer Learning used as a classifier which result with good performance accuracies are from 54.5% to 65.5% in case of using DWT and 53.6% to 62.4% in case of using PSD. Moreover, Cross-validation were used to ensure the validity of data by splitting in three categories train, validation and test data and by this there was no replications. 
Supervisor : Dr. Hatem Rmili 
Thesis Type : Master Thesis 
Publishing Year : 1442 AH
2020 AD 
Co-Supervisor : Dr. Mohammed Abdulaal 
Added Date : Thursday, August 27, 2020 

Researchers

Researcher Name (Arabic)Researcher Name (English)Researcher TypeDr GradeEmail
ماهر عبدالرحمن الجهنيAljohani, Maher AbdulrahmanResearcherMaster 

Files

File NameTypeDescription
 46705.pdf pdf 

Back To Researches Page