(For USM Staff/Student Only)

EngLib USM > Ω School of Electrical & Electronic Engineering >

Fusion of singular value decomposition and band limited phase only correlation

Fusion of singular value decomposition and band limited phase only correlation / Chee Kok Wei
Dalam tesis ini, gabungan dua algoritma termasuk pengecaman urat jari berdasarkan nilai penguraian tunggal (SVD) dengan minutia sepadan dan pengecaman dengan saluran terhad fasa sahaja korelasi (BLPOC) dicadangkan untuk meningkatkan ketepatan. Kaedah pengabungan yang diguna adalah mesin sokongan vektor (SVM). Dalam kebanyakan kes berasaskan kaedah algoritma tunggal, terdapat banyak batasan bagi setiap daripada mereka seperti terdedah kepada masalah putaran atau translasi, gambar yang tidak jelas atau pengambaran urat yang terjejas dalam gambar urat jari. Mesin sokongan vektor digunakan untuk melaksanakan gabungan dalam tahap skor untuk kedua-dua algoritma. Kaedah SVD dari domain masa manakala kaedah BLPOC adalah dari domain frekuensi. Dengan sifat-sifat ini, prestasi dan ketepatan hasil gabungan boleh dipertingkatkan disebabkan oleh masalah dalam domain masa mungkin dapat diselesaikan oleh komponen dalam domain frekuensi jadi masalah untuk manamana domain tidak akan memberi kesan kepada kedua-dua algoritma. Eksperimen telah diuji dengan pangkalan data FV-USM yang terdiri daripada 123 sukarelawan. SVM menunjukan keputusan dengan EER 0.04% berbanding dengan kaedah SVD dengan EER sebanyak 13% dan kaedah BLPOC dengan EER sebanyak 2.7%. Oleh itu, gabungan teknik SVD dan BLPOC dengan SVM menunjukkan keputusan yang paling baik dibandingkan dengan cara yang lain dalam pengecaman urat jari. _______________________________________________________________________________________________________ In this thesis, fusion of two algorithms including finger vein recognition based on singular value decomposition (SVD) with minutiae matching and finger vein recognition with band limited phase only correlation (BLPOC) is proposed to increase the accuracy. The method for fusion is support vector machine (SVM). In most of the cases for single algorithm based method, there are many limitations for each of them such as vulnerable to rotational or translational problems, noise and occlusion in the finger vein image. Support vector machine is used to perform score level fusion for both of the algorithms. SVD method is from spatial domain while the BLPOC method is from frequency domain. With these properties, the performance and accuracy of the fusion result can be greatly enhanced due to difficulty in spatial domain can be solved by frequency domain component so that difficulty in either domain will not affect both algorithms. The experiment was conducted using FV-USM database which based on 123 volunteers. SVM achieved a result with EER of 0.04% compare to SVD method with EER of 13% and BLPOC method with EER of 2.7%. Therefore, fusion of SVD and BLPOC method with SVM perform the best in finger vein recognition compared the other methods.
Contributor(s):
Chee Kok Wei - Author
Primary Item Type:
Final Year Project
Identifiers:
Accession Number : 875008004
Language:
English
Subject Keywords:
algorithms; (SVD); (BLPOC)
First presented to the public:
6/1/2015
Original Publication Date:
3/21/2019
Previously Published By:
Universiti Sains Malaysia
Place Of Publication:
School of Electrical & Electronic Engineering
Citation:
Extents:
Number of Pages - 94
License Grantor / Date Granted:
  / ( View License )
Date Deposited
2019-03-21 16:51:53.563
Submitter:
Mohd Jasnizam Mohd Salleh

All Versions

Thumbnail Name Version Created Date
Fusion of singular value decomposition and band limited phase only correlation1 2019-03-21 16:51:53.563