Effective features and machine learning methods for human recognition based on multi-biometric systems

Le résumé fourni par la source

Biometrics are fundamental to a wide range of technologies that require credible authentication approach to approve personal identification. This thesis aims to identify effective features and machine learning methods for human recognition based on multiple biometrics and produce the sufficient combination of single biometric systems suitable in specific applications for identification purposes. For example, banking systems which use multi-biometric authentication for login procedures and the police and criminal evidence applications. This thesis goes through general ideas of the most common biometrics that used for a personal identification and their application areas. It has been focusing on two well-known linear subspace-learning techniques that have become the most popular techniques for face recognition; PCA and LDA. Different face classification techniques have been presented including supervised and unsupervised learning methods. The research focuses on assessing vision system performance and different databases that are suitable for biometric research. This research has made a number of contributions. Firstly improving the Viola-Jones face detection performance by using the brightness channel in HSV And HSL color spaces. Distance similarity measures have been compared for PCA- and LDA-based face, ear and palm biometrics. The face and ear recognition performance using SVM based on PCA and SVM based on a combination of PCA and LDA techniques have been compared with the PCA and LDA techniques based on distance similarity measures. Face, ear, palmprint, eye, and hand biometric recognition has been applied using three Deep and Shallow Convolutional Neural Networks (GoogleNET, VGG16, ResNET). An implementation of a person identification system fusing different combinations of biometric modalities; face, ear, eye, hand, and palmprint at score level has been examined. Comparison of these combinations has been employed to assess the performance. Finally, different sizes of training/testing set are examined to achieve high recognition performance.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.