Enhanced chicken swarm optimization-tuned convolutional neural network for fingerprint-based ethnicity identification

Abstract

Identification of human being based on fingerprints have proven to be highly reliable. Researches have been done on fingerprint ethnicity identification, which are characterized with high false positive rate and high recognition time. This research developed Fingerprint Identification System with an enhanced CSO combined with CNN for better fingerprint ethnicity identification. One thousand two hundred (1200) subjects’ fingerprint images were acquired from three major ethnic groups in Nigeria (Yoruba, Igbo and Hausa) with equal ratio of male to female between the ages of 17–50 years, using Secugen Hamster Plus Fingerprint Scanner; Six hundred (600) acquired subjects’ fingerprints were augmented and used for training while the remaining 50% were used for testing. The raw images were pre-processed; CNN hyperparameters were tuned using CSO and CSO enhanced with Chaotic theory. The implementation was done using MATLAB R2023a software. The performance of the ICSO-CNN was evaluated and compare with CNN and CSO-CNN at a benchmark of 0.75 threshold value using, False Positive Rate (FPR) and Recognition Time (RT). The FPR and RT using CNN for Yoruba, Igbo and Hausa were 3.5% and 60.54s; 3.75% and 59.31s and 4% and 58.04s, respectively. The FPR and RT using CSO-CNN for Yoruba, Igbo and Hausa were 2.25% and 49.87s; 2.5% and 48.82s and 2% and 49.79s, respectively, while the corresponding values for the enhanced CSO-CNN were 1.75% and 36.64s, 2% and 38.56s and 1.5% and 39.35s. The developed fingerprint-based ethnicity identification system gave an improved identification performance over CNN and CSO-CNN. The developed ICSO-CNN can be used by security agencies for proper identification of criminals based on ethnicity.