An intelligent system for automatic fingerprint identification using feature fusion by Gabor filter and deep learning. (October 2021)

• Record Type: Journal Article
• Authors: Nur-A-Alam, ; Ahsan, M.; Based, M.A.; Haider, J.; Kowalski, M.
• Is Part Of: Computers & electrical engineering. Volume 95(2021)
• Publication Date: 2021-10
• Subjects: Fingerprint authentication -- Convolutional neural network (CNN) -- Gabor filter -- Principle component analysis (PCA) -- Deep learning; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2021.107387
• Languages: English
• ISSNs: 0045-7906

Abstract

Abstract: This paper introduces an intelligent computational approach to automatically authenticate fingerprint for personal identification and verification. The feature vector is formed using combined features obtained from Gabor filtering technique and deep learning technique such as Convolutional Neural Network (CNN). Principle Component Analysis (PCA) has been performed on the feature vectors to reduce the overfitting problems in order to make the classification results more accurate and reliable. A multiclass classifier has been trained using the extracted features. Experiments performed using standard public databases demonstrated that the proposed approach showed better performance with regard to accuracy (99.87%) compared to the more recent classification techniques such as Support Vector Machine (97.86%) or Random Forest (95.47%). However, the proposed method also showed higher accuracy compared to other validation approaches such as K-fold (98.89%) and generalization (97.75%). Furthermore, these results were supported by confusion matrix results where only 10 failures were found when tested with 5000 images.