Kristian, Yosi (2018) Analisa Citra Wajah Bayi Untuk Deteksi Nyeri Dan Tangis Menggunakan Multi Stage Classification Dan Deep Learning. Doctoral thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
2213301008-Disertation.pdf - Accepted Version Download (4MB) | Preview |
Abstract
Bayi belum dapat menginformasikan rasa nyeri yang mereka alami, karena itu bayi menangis saat mereka mengalami nyeri. Dengan semakin berkembangnya teknologi visi komputer, beberapa tahun terakhir muncul beberapa penelitian yang mencoba mengenali nyeri pada tangis bayi memanfaatkan machine learning dan pengolahan citra. Dalam disertasi ini diteliti tentang analisa pemanfaatan handcrafted feature dan feature learning dengan bantuan berbagai machine learning untuk mengatasi masalah klasifikasi nyeri pada wajah bayi. Pada penelitian menggunakan handcrafted feature dicari kombinasi fitur geometri dan tekstur yang paling optimal untuk dapat mengklasifikasikan nyeri dengan bantuan multi stage classification menggunakan SVM. Sedangkan pada penelitian dengan feature learning digunakan arsitektur deep learning dengan tipe Deep Convolution Neural Network (DCNN) Autoencoder dan Long-Short Term Memory (LSTM) Network. Dari penelitian yang dilakukan performa handcrafted feature ditambah dua tahap klasifikasi cukup seimbang jika dibandingkan dengan performa deep learning. Keunggulan penelitian deep learning dengan memanfaatkan autoencoder dan LSTM adalah kemampuan mengolah video atau sequence gambar.
===========================================================================================
Babies are yet able to inform the pain they are experiencing, s
o babies cry
when they experience pain. With the development of computer vision technology,
the last few years have emerged several studies that try to recognize the pain in
babies faces using machine learning and image processing.
In this dissertation we
try to analyze the use of handcrafted features and
feature learning with the help of machine learning to tackle the infant facial pain
classification problem. In the research using handcrafted features we found the most
optimal combination for geometrical
and textural features to classify pain in infant
with the help of multi stage classification using SVM. While in the research with
feature learning we use deep learning architecture with Deep Convolution Neural
Network (DCNN) Autoencoder and Long
-
Short Ter
m Memory (LSTM) Network.
From the research conducted, the performance of handcrafted feature plus
two stages of classification is quite balanced when compared with the performance
of the one using deep learning. The superiority of our deep learning researc
h is the
ability to process video or picture sequences because the usage of LSTM.
| Item Type: | Thesis (Doctoral) |
|---|---|
| Additional Information: | RDE 006.42 Kri a |
| Uncontrolled Keywords: | Infant facial cry detection, infant facial pain classification, geometrical feature, textural feature, feature learning, deep learning,Deteksi tangis bayi, klasifikasi nyeri pada bayi, fitur geometri, fitur tekstur, feature learning, deep learning |
| Subjects: | T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7882.P3 Pattern recognition systems |
| Divisions: | Faculty of Electrical Technology > Electrical Engineering > 20001-(S3) PhD Thesis |
| Depositing User: | Yosi Kristian |
| Date Deposited: | 01 Oct 2020 07:09 |
| Last Modified: | 01 Oct 2020 07:09 |
| URI: | http://repository.its.ac.id/id/eprint/59669 |
Actions (login required)
 |
View Item |