Integrated Different Fingerprint Identification and Classification Systems based Deep Learning

Fingerprint classification plays a fundamental role in fingerprint identification systems because it dramatically reduces the time spent identifying an individual. This classification can be based on gender, as fingerprints can be used to identify and differentiate between genders. The gender classification based on fingerprints is a crucial step in different applications such as airports, banks, explosions and natural disasters, forensic anthropology for identifying the gender of the offender and reducing the list of suspects' searches. It helps in efficiently analyzing the data and helping to sort the data. This study proposes a new idea based on integrating the Wiener filter, multi-level Histogram techniques with three Convolutional Neural Networks for classifying and identifying fingerprints in terms of gender, hand, and fingers utilizing a publicly accessible benchmark of a SOCOFing dataset. The fingerprint dataset augmentation and pre-processing are applied for expanding and enhancing the dataset. Then, the features were extracted using convolutional neural networks followed by Softmax as a classifier. It is worth mentioning that the dropout layer is used to prevent overfitting. Evaluation results have proven the high performance of accuracies: 99.96% for identifying the gender (Male or Female), 99.93% for identifying the hand type (Right or Left), and 99.90% for identifying the fingers.