S. Akbar
Bandung Institute of Technology

Published : 2 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 2 Documents
Search

Fingerprint Direct-Access Strategy Using Local-Star-Structure-based Discriminator Features: A Comparison Study G. Indrawan; S. Akbar; B. Sitohang
International Journal of Electrical and Computer Engineering (IJECE) Vol 4, No 5: October 2014
Publisher : Institute of Advanced Engineering and Science

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (1381.794 KB)

Abstract

This paper describes a comparison study of the proposed fingerprint direct-access strategy using local-star-topology-based discriminator features, including internal comparison among different concerned configurations, and external comparison to the other strategies. Through careful minutiae-based feature extraction, hashing-based indexing-retrieval mechanism, variable-threshold-on-score-ratio-based candidate-list reduction technique, and hill-climbing learning process, this strategy was considered promising, as confirmed by the experiment results. For particular aspect of external accuracy comparison, this strategy outperformed the others over three public data sets, i.e. up to Penetration Rate (PR) 5%, it consistently gave lower Error Rate (ER). By taking sample at PR 5%, this strategy produced ER 4%, 10%, and 1% on FVC2000 DB2A, FVC2000 DB3A, and FVC2002 DB1A, respectively. Another perspective if accuracy performance was based on area under curve of graph ER and PR, this strategy neither is the best nor the worst strategy on FVC2000 DB2A and FVC2000 DB3A, while on FVC2002 DB1A it outperfomed the others and even it gave impressive results for index created by three impressions per finger (with or without NT) by ideal step down curve where PR equal to 1% can always be maintained for smaller ER.DOI:http://dx.doi.org/10.11591/ijece.v4i5.6589
A Multi-Threaded Fingerprint Direct-Access Strategy Using Local-Star-Structure-based Discriminator Features G. Indrawan; B. Sitohang; S. Akbar; A. S. Nugroho
Indonesian Journal of Electrical Engineering and Computer Science Vol 12, No 5: May 2014
Publisher : Institute of Advanced Engineering and Science

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Taking advantage of multi-thread technology provided by multi-core in single processor, this paper describes multi-thread implementation on fingerprint direct-access strategy using local-star-topology-based discriminator features. Multi-thread was applied for data partitioning on hashing add-lookup process, and for work partitioning on similarity score computation. Efficiency of the implementation was achieved, as confirmed by the results of experiments. Using quad-thread of dual-core single processor on last three largest databases on experiment, multi-thread implementation can reduce searching time at relatively constant value around 300 ms compare to its single-thread implementation. This result achieved with relatively same accuracy trend and has significant improvement by considering millisecond order of searching process on large scale data. DOI : http://dx.doi.org/10.11591/telkomnika.v12i5.5208