Forensic Data Analysis as A Problem of Pattern Recognition ? An Empirical Study

A latent fingerprint, a drop of blood, hair is the successful forensic clues used to catch criminals for decades. The criminals are identified by matching the clues with their physiological characteristics. Pattern Recognition is a process of matching a sample to determine the category/group/class to which it belongs to. This work concentrates on classification of fingerprints. We extracted features of fingerprints using co-occurrence matrices and compared the performance of classifiers KNN and SVM. Experiments are conducted on 384 fingerprints and observed that SVM performs well when compared to KNN.