A Set of Convolutional Neural Networks for Person Identification with Different Biometrics
Biometric Identification is one of the most intriguing fields of research in the broad area of pattern recognition and computer vision and it finds wide applications in government institutions, forensic research and crime investigation, employee and student attendance roll, restricted access to classified information, and in various public as well as private enterprises. Human face is an age-old biometric and recognition is based on the unique facial structure and symmetry of an individual. Fingerprint is a very important biometric for forensic research as it can be collected easily from the crime scene and also every individual has unique patterns on every finger. Human iris has unique patterns for any particular individual and thus iris verification is gaining ground as an important person identity authentication method. In the present paper these three biometrics have been chosen because these are commonly available and most frequently used. The present Convolution Neural Network (CNN) based person identification system with different biometrics is simple but efficient, effective, and fast. The performance evaluation of the system in terms of accuracy, precision, recall and f-score derived from confusion matrix together with training and testing time is moderate, appreciable and affordable. Keywords - Pattern Recognition; Convolution Neural Network; Biometric; Holdout Method; Confusion Matrix; Face; Fingerprint; Iris; Accuracy; Precision; Recall; F-Score.