DEVELOPMENT OF A BIOMETRIC (FINGERPRINT) STUDENT ATTENDANCE SYSTEM USING GLOWWORM SWARM OPTIMIZATION

ABSTRACT The development of automatic fingerprint recognition systems has been an active area of research since the 1960s, initially achieving significant success in forensic applications. These advancements have since extended to civilian uses. While various machine learning techniques have been applied to address challenges in this domain, none have consistently delivered optimal results. This project aims to design a fingerprint-based attendance system employing Glowworm Swarm Optimization (GSO) for feature selection and Back Propagation Neural Network (BPNN) for classification. A total of 432 fingerprint samples were collected from students of LAUTECH and processed through multiple stages. Feature extraction was performed using a Gabor filter, while GSO was employed for selecting salient features, and classification was carried out using BPNN. The dataset was split into 60% for training and 40% for testing. Simulations were conducted in MATLAB 2016a, with performance evaluated using metrics such as recognition accuracy, false positive rate (FPR), sensitivity, specificity, and average recognition time. The developed system achieved a recognition accuracy of 98.38%, an FPR of 1.85%, sensitivity of 98.46%, and specificity of 98.38% with a threshold of 0.75 in an average computation time of 83.08ms. By comparison, the developed system without Gabor filter attained an FPR of 6.48%, sensitivity of 96.91%, specificity of 93.52%, and accuracy of 96.06% in 92.16ms of computation time.