Machine Learning System for Recognition and Classification of Overlapped Fingerprints

Abstract

Latent fingerprints were found frequently in criminal investigations. Thus, Overlapped Fingerprint Recognition (OFR) technology plays key role in many applications. The OFR technology is a relatively new area which is a challenging and critical area of research work. The conventional methods are struggles in achieving high accuracy due to improper features. Thus, this article focused on implementation of OFR technology with multiple descriptors based modified dimensionality reduction mechanism. The proposed OFR is developed with gradient variation approach by using Kirsch edge detection to overcome the problems of conventional approaches. The dimension of the extracted feature space is reduced using the Kernel Principal Component Analysis (KPCA) method. Finally, Support Vector Machine (SVM) classifier is applied to classify the overlapped region of test image by comparing with the training database. Simulation results shows that the proposed method increases accuracy, specificity and sensitivity as compared to the existing methods.