|
English
|
正體中文
|
简体中文
|
Items with full text/Total items : 62830/95882 (66%)
Visitors : 4030760
Online Users : 975
|
|
|
Loading...
|
Please use this identifier to cite or link to this item:
https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/122622
|
Title: | Classification of Music-Induced Mental States Using Convolutional Neural Networks for an EEG Study |
Authors: | Cheah, Kit Hwa;Nisar, Humaira;Tsai, Chi-Yi |
Keywords: | Electroencephalogram (EEG);Convolutional neural network (CNN);Music and binaural beats |
Date: | 2020-11-24 |
Issue Date: | 2022-03-26 12:10:59 (UTC+8) |
Abstract: | Electroencephalogram (EEG) is the brain signal acquired through multiple channels and is packed with useful information for the mental state recognition. EEG has wide applicability in the field of medicine (e.g. pre-disease risk estimation, disease characterization, prognosis, treatment monitoring), psycho-physiological research (e.g., affective state classification, stress assessment, alertness monitoring, sleep stage identification), human–computer interaction (e.g. thought typing, prosthetic limb control), and many other areas. However, manual EEG feature selection is time-consuming and challenging to fully make use of the relevant information embedded in the EEG signals. Deep learning (DL), while enabling high hierarchical abstract representation of complex data, has been strongly indicated by recent research works to be generally outperforming the manual EEG feature extraction algorithms and classical classifiers. In this study, different neural network architectures have been constructed for binary classification of an EEG using a dataset that showed no significant statistical difference between its two data categories (baseline resting EEG and EEG while listening to music) with traditional EEG feature extraction methods. The two-convolutional-path Convolutional Neural Network (CNN) model examined in this study shows higher validation accuracy (75 ± 1%) than the single-convolutional-path CNN model examined which had achieved accuracy of 71.5 ± 2% using the same EEG dataset. The effects of different architectural components and model training hyperparameters on the models’ validation performance are also studied and presented. |
Relation: | Proceedings of the 12th National Technical Seminar on Unmanned System Technology 2020, p. 383-401 |
DOI: | 10.1007/978-981-16-2406-3_30 |
Appears in Collections: | [Graduate Institute & Department of Electrical Engineering] Proceeding
|
Files in This Item:
File |
Description |
Size | Format | |
index.html | | 0Kb | HTML | 37 | View/Open |
|
All items in 機構典藏 are protected by copyright, with all rights reserved.
|