Paternity Test from Offspring Fingerprint Using Parallel Convolutional Network

The research of Fingerprint Ridge Structure (FRS), conducted by early pioneers in the field of dermatoglyphics, demonstrated a clear correlation between the size, shape, and spacing of the ridges and the inheritance of a fingerprint pattern. This article proposes a novel approach to identify paternity or hereditary relationship among parent and child with the help of fingerprint using deep network learning. Fingerprint image of five fingers of right hand from father, mother and child are acquired from 100 families to prepare a dataset for training and testing of the proposed network. A parallel deep convolution network is developed and trained with the dataset for heredity testing. The network achieve a 87.94\% training accuracy and 82.23\% validation accuracy. Based on the outcome of the proposed method, we claim its possibility to identify hereditical relationship using fingerprint and the proposed solution may also be useful in various clinical and genetically correlated studies and court of law. HighlightsA deep learning approach to determine paternity relation from offspring fingerprint image.A parallel deep learning network that accepts parent and offspring fingerprint image as input and verifies the paternity relationship between the fingerprints.The deep neural network model is too lightweight to implement in any device like mobile phone or tiny biometric authentication device.The network can able to produce satisfactory outcome even if trained in a limited time on any dataset.