Design and Development of a Biometric Recognition System to Uniquely Identify Sexual Assault Offenders

Abstract: This study designed and developed a biometric recognition system to uniquely identify sexual assault offenders in Zambia for lack of empirical evidence for convictions. A simulation-based study was conducted using 6000 fingerprint images from the SOCOFing database. The fingerprint recognition algorithm showed 94.7% accuracy with a 2.3% False Acceptance Rate (FAR), 3.0% False Rejection Rate (FRR) and Equal Error Rate (EER) of 2.6%. Facial recognition system presented an accuracy of 89.5% (FAR = 3.8%, FRR = 6.7%). The multimodal fusion of fingerprint and facial recognition increased the system's accuracy to 97.2%. Database performance test results denote average query response time, being 0.3 seconds for a single-fingerprint match among ten thousand records, while usability test scores graded by fifty law enforcement agents showed a System Usability Scale score of 82, indicating that the biometric detection system works in safer user environment demonstrating sutures with high detection rate ranging results from [98,99]. The proposed systems achieved an overall indiscriminateness up to detecting spoofs by reaching a spoof detection usable rate of 99%.