EKSTRAKSI CIRI BERBASIS FILTER BANK GABOR SEBAGAI SISTEM VERIFIKASI SIDIK JARI

• Main Author: Adi, Kusworo
• Format: Theses
• Language: Indonesia
• Subjects: Teknik (Rekayasa, enjinering dan kegiatan berkaitan)
• Online Access: https://digilib.itb.ac.id/gdl/view/5284
• Institution: Institut Teknologi Bandung

<b>Abstract :</b><p align=\"justify\"> <br /> Biometric is a development of basic method of identification using human natural characteristics as its basis One of the biometric system that is often used is fingerpint.The existing fingerprint matching system is based on minutiae algorithm.However,minutiae algorithm still has a problem especially in local ridge which has not been characterized completely. This weakness will appear in the matching process between two fingerprints that contain difference in number of unregistered minutiae points.<p align=\"justify\"> <br /> To solve this problem an algorithm based on texture which can extract local and global information of fingerprint as feature vector is proposed Texture analysis calculates feature from fingerprint pattern using multichannel filtering such as Gabor filter. Multichannel filtering is capable of resembling characteristic of human visual system which is sensitive to orientation and frequency.<p align=\"justify\"> <br /> Fingerprint feature is extracted by average absolute deviation (AAD) method as feature vector which is used in the matching process between two fingerprints. This process uses Euclidean distance by giving total score on a scale from 0 to 50.T he resemblance between the two fingerprints will be seen in the number of high matching score.<p align=\"justify\"> <br /> In this testing system 150 fingerprints data were taken using fingerprint scanner,Finger TIPT 11100 MF I V2.0, made by Infenion. From 150 fingerprints data,14 fingerprints (9.33%) were rejected and 59 fingerprints (90.67%) were accepted by the system. The rejection was caused by fingerprint damage which happened in the process of capturing images.In the verification system testing of fingerprint, we obtained error value of system 4,8% in value of threshold 37. While value of receiver Operating Characteristic(ROC) was in Genuine Acceptance Rate (GAR) area from 88% to 98% with value of False Acceptance Rate (FAR) from 0,0% to 8,5%. The speed of fingerprint verification process was influenced by the size of Gabor filter kernel and the processor which was used From the testing result, we conclude that this system can operate well. <br /> <br /> <br />