Classification Of Gender Using Global Level Features In Fingerprint For Malaysian Population

• Main Author: Abdullah, Siti Fairuz
• Format: Thesis
• Language: English; English
• Published: UTeM 2016
• Subjects: T Technology (General); TK Electrical engineering. Electronics Nuclear engineering
• Online Access: http://eprints.utem.edu.my/id/eprint/18564/1/Classification%20Of%20Gender%20Using%20Global%20Level%20Features%20In%20Fingerprint%20For%20Malaysian%20Population%2024%20Pages.pdf; http://eprints.utem.edu.my/id/eprint/18564/2/Classification%20Of%20Gender%20Using%20Global%20Level%20Features%20In%20Fingerprint%20For%20Malaysian%20Population.pdf; http://eprints.utem.edu.my/id/eprint/18564/; https://plh.utem.edu.my/cgi-bin/koha/opac-detail.pl?biblionumber=100896

A new approach of algorithm based on the Mark Acree’s theory, focusing on fingerprint global extracted features is proposed and implemented for enhancing gender classification method. This proposed method can automatically execute the ridge calculation process from the 25mm2 fingerprint and enhance the forensic gender classification process. In this study, a relationship between fingerprint global features and a gender of person in Malaysian population is also explored, enhanced and improved by exploiting another five additional fingerprint features. A sample of 3000 fingerprints from 300 respondents of random selection are carefully taken before any relationship can be determined. For the classification part, five extracted features of the fingerprint are used which are Ridge Density (RD), Mean Ridge Count (RC), Ridge Thickness to Valley Thickness Ratio (RTVTR), White Lines Count (WLC) and Mean Pattern Types (PT). Two classification approaches which are the descriptive statistical and data mining are used in order to examine the classification of the gender by using the five extracted features. For data mining classification part, there are four popular machine learning classifiers used which are Bayesian Net.work (Bayes Net.), Multilayer Perceptron Neural Network (MLPNN), K-Nearest Neighbor (KNN) and Support Vector Machine (SVM). These four classifiers are used in the data mining task with five test cases each in order to find the accuracies of the gender classification. The accuracy of the results from the proposed method is compared to the Acree Method is shown in terms of relative error. For statistical approach using Ridge Density (RD), the relative error is 3.7% for male respondent and 4.1% for female respondent. Meanwhile, the overall performance of the result from the proposed method achieved more than 90% classification rate for all the classifiers. SVM emerges as the best classifier for all the different cases in order to classify the gender using the results from the proposed method.