TY - GEN
T1 - Enhancing Face Recognition Accuracy Using the ED-FFP Extraction Method and Ensemble Learning for Forensics and Cyber Security
AU - Virmani, Pranav
AU - Prabhu, Srikanth
AU - S, Ramya
N1 - Publisher Copyright:
© 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
PY - 2023
Y1 - 2023
N2 - Currently, Face Recognition is the most used biometric to determine an individual’s identity due to its natural and unobtrusive nature. This study proposes face recognition and verification of two-dimensional(2D) images by a feature extraction algorithm which involves identifying anthropological facial feature points and calculating the Euclidean Distance (ED) between these points (ED-FFP) as these distances, if required can be used as an objective measure during trials in courts. These measurements are then used as inputs for various classification methods, including Logistic Regression Classifier (LR), Decision Tree (DT), Naive Bayes (NB), and two other classifiers using the Ensemble Learning Model. The method was tested on 2D face image databases (Caltech, Yale, and ORL) and found to be more efficient and accurate for face recognition than other methods, with a maximum accuracy of 85% for predicting distinct faces using the Decision Tree classifier model. The ensemble learning model also had an accuracy of 85%, which could potentially be improved by using more photos for comparison. In future work, the method could be applied to 3D images, which is currently an open challenge in the field.
AB - Currently, Face Recognition is the most used biometric to determine an individual’s identity due to its natural and unobtrusive nature. This study proposes face recognition and verification of two-dimensional(2D) images by a feature extraction algorithm which involves identifying anthropological facial feature points and calculating the Euclidean Distance (ED) between these points (ED-FFP) as these distances, if required can be used as an objective measure during trials in courts. These measurements are then used as inputs for various classification methods, including Logistic Regression Classifier (LR), Decision Tree (DT), Naive Bayes (NB), and two other classifiers using the Ensemble Learning Model. The method was tested on 2D face image databases (Caltech, Yale, and ORL) and found to be more efficient and accurate for face recognition than other methods, with a maximum accuracy of 85% for predicting distinct faces using the Decision Tree classifier model. The ensemble learning model also had an accuracy of 85%, which could potentially be improved by using more photos for comparison. In future work, the method could be applied to 3D images, which is currently an open challenge in the field.
UR - http://www.scopus.com/inward/record.url?scp=85161193838&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85161193838&partnerID=8YFLogxK
U2 - 10.1007/978-981-99-2264-2_11
DO - 10.1007/978-981-99-2264-2_11
M3 - Conference contribution
AN - SCOPUS:85161193838
SN - 9789819922635
T3 - Communications in Computer and Information Science
SP - 130
EP - 142
BT - Applications and Techniques in Information Security - 13th International Conference, ATIS 2022, Revised Selected Papers
A2 - Prabhu, Srikanth
A2 - Pokhrel, Shiva Raj
A2 - Li, Gang
PB - Springer Science and Business Media Deutschland GmbH
T2 - 13th International Conference on Applications and Techniques in Information Security, ATIS 2022
Y2 - 30 December 2022 through 31 December 2022
ER -