Deteksi Ekspresi Wajah Menggunakan Fitur Gabor dan Haar Wavelet

Primasiwi, Claudia (2018) Deteksi Ekspresi Wajah Menggunakan Fitur Gabor dan Haar Wavelet. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.

[img] Text
5112100140-Undergraduate_Theses.pdf - Published Version
Restricted to Repository staff only

Download (3MB) | Request a copy

Abstract

Manusia bersosialisasi secara verbal dan non-verbal. Salah satu bahasa non-verbal yang sering digunakan manusia untuk berinteraksi adalah ekspresi wajah. Sistem pengenalan ekspresi wajah manusia dilakukan untuk mengembangkan interaksi yang natural antara manusia dan komputer. Penelitian ini dibangun dalam tiga tahap: deteksi area wajah dan pre-processing, ekstraksi fitur, dan klasifikasi. Pendeteksian area wajah dilakukan dengan algoritma Viola-Jones. Pre-processing dilakukan menggunakan histogram equalization. Selanjutnya dilakukan ekstraksi fitur Gabor dan fitur Haar. Untuk melakukan klasifikasi, digunakan SVM One-vs-All. Fitur Gabor digunakan karena invarian terhadap rotasi, penskalaan, dan translasi. Sedangkan fitur Haar digunakan karena komputasinya yang efisien dan efektif dalam merepresentasikan sinyal dalam dimensi rendah dan tetap mempertahankan energinya. Pada penelitian ini, digunakan kombinasi fitur Gabor dan Haar untuk dibandingkan dan digabungkan dengan fitur Landmark yang telah disediakan di database. Berdasarkan hasil uji coba, fitur Gabor dan Haar memiliki akurasi sebesar 92% dengan kelas terbaik berupa ekspresi senang. ============== Humans socialize verbally and non-verbally. One of the non-verbal interaction that frequently used is facial expression. Facial expression recognition is developed to create a natural interaction between human and computer. This research is built in three phases: face detection and pre-processing, feature extraction, and classification. Facial area is detected using Viola-Jones algorithm and pre-processed by histogram equalization. Next, features are extracted using Gabor and Haar wavelet. Finally, the classification is done using One-vs-All SVM. Gabor is selected due to its invariance to rotation, scaling, and translation while Haar feature is selected because of its efficiency and effectivity to compute in low dimensional signal and preserve its energy. In this research, Gabor and Haar features are used to be compared and to be combined with Landmark features which already enclosed to the database. The best features are Gabor and Landmark features resulting in 92% accuracy with happy as the best class.

Item Type: Thesis (Undergraduate)
Uncontrolled Keywords: Facial Expression; Gabor; Haar Wavelet; Viola-Jones; SVM; Ekspresi Wajah
Subjects: Q Science > QA Mathematics > QA403.3 Wavelets (Mathematics)
T Technology > T Technology (General)
T Technology > TA Engineering (General). Civil engineering (General) > TA1650 Face recognition
Divisions: Faculty of Information Technology > Informatics Engineering > (S1) Undergraduate Theses
Depositing User: Claudia Primasiwi
Date Deposited: 06 Mar 2018 03:10
Last Modified: 06 Mar 2018 03:10
URI: http://repository.its.ac.id/id/eprint/50384

Actions (login required)

View Item View Item