Klasifikasi Emosi pada Sinyal EEG dengan CNN Berdasarkan Fitur Brain Connectivity

Oktaviany, Dewi (2025) Klasifikasi Emosi pada Sinyal EEG dengan CNN Berdasarkan Fitur Brain Connectivity. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5024211041_Undergraduate_Thesis.pdf Restricted to Repository staff only Download (20MB) | Request a copy

Abstract

Pemanfaatan fitur brain connectivity pada sinyal EEG menawarkan representasi keterhubungan fungsional otak yang relevan untuk klasifikasi emosi. Fitur korelasi dan koherensi antar-kanal EEG pada berbagai pita frekuensi diekstraksi dari data EEG DREAMER dengan 14 kanal dari 23 partisipan yang menonton klip video pembangkit emosi. Data EEG diproses melalui segmentasi waktu, filterisasi, lalu dihitung matriks konektivitas sebagai fitur. Fitur-fitur ini digabungkan, dinormalisasi, dan digunakan sebagai input dua arsitektur CNN. Hasil eksperimen menunjukkan bahwa kedua model CNN mampu mengklasifikasikan emosi dalam empat kondisi (HALV, HAHV, LAHV, LALV) dengan performa yang baik. Model pertama mencapai akurasi 82% dengan konvergensi stabil, sedangkan model kedua mencapai akurasi lebih tinggi, 94%, namun menunjukkan potensi overfitting. Hasil ini menegaskan bahwa integrasi brain connectivity dalam CNN dapat meningkatkan akurasi klasifikasi emosi berbasis EEG.
=========================================================================================================================================
Utilizing brain connectivity features in EEG signals provides a representation of the brain’s functional interconnections that is relevant for emotion classification. Correlation and coherence features between EEG channels across multiple frequency bands were extracted from the DREAMER EEG dataset,which contains recordings from 23 participant swatching emotion evoking video clips. The EEG data underwent time segmentation, filtering, and computation of connectivity matrices as features. These features were combined, normalized, and used as in put to two different CNN architectures. Experimental results showed that both CNN models successfully classified emotions into four conditions (HALV, HAHV, LAHV, LALV) with good performance. The first model achieved 82% accuracy with stable convergence, while the second model achieved a higher accuracy of 94%, albeit with signs of overfitting. These findings confirm that integrating brain connectivity features into CNNs can enhance the accuracy of EEG-based emotion classification.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	EEG, Emosi, Brain Connectivity, CNN, Korelasi, Koherensi, EEG, Emotion, Brain Connectivity, Correlation, Coherence, CNN.
Subjects:	Q Science > QP Physiology Q Science > QP Physiology > Q376.5 Electroencephalography (EEG) T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Electrical Technology > Computer Engineering > 90243-(S1) Undergraduate Thesis
Depositing User:	Dewi Oktaviany
Date Deposited:	04 Aug 2025 04:33
Last Modified:	04 Aug 2025 04:33
URI:	http://repository.its.ac.id/id/eprint/125984

Actions (login required)

View Item