Mariella Corvino

Face Based Recognition Algorithms: A First Step Toward a Metrological Characterization

Face-based recognition systems have been increasingly used in many different applications in todays society, starting from surveillance and access control to the authentication for banking activities. Therefore, in the last few years an increasing interest in the performance characterization and improvement of such systems can be found in the scientific community. Most of the methodologies for testing the performance of such systems are based on the evaluation of recognition reliability indexes that are generally related to the probability of a false positive and/or of a false negative.

Estimation of influence quantities in face recognition

Nowadays, the matter of uncertainty in face recognition based biometric systems is a relevant issue for the scientific community. This is due to the even more increasing deployment of such systems in critical applications as safety, security and access control, to cite a few. In this context, the authors are engaged in the design of general methods for uncertainty modeling and evaluation aimed at realizing face recognition based biometric systems with built-in uncertainty evaluation capability. In this way, the output of a recognition system will not be the identity of the observed subject, but a confidence level for each possible subject. In previous papers the authors have identified the quantities of influence and have proposed a suitable uncertainty model. Core of the proposed model is the knowledge of the value assumed by the quantities of influence with respect to the corresponding values achieved in suitable reference conditions. This paper mainly analyzes these measurement issues, a fundamental step toward the development of such systems with built-in uncertainty evaluation capability. First results show a good agreement between statistical indicators and a priori estimations achieved with the proposed method.

A Proposal for the Management of the Measurement Uncertainty in Classification and Recognition Problems

This paper proposes a new approach to classification and recognition problems. It considers the measurement uncertainty affecting input data to improve the overall effectiveness of this type of process. The proposed method is based on an effective probabilistic approach for the evaluation of the confidence level of system outputs and the suitable use of related information to improve the performance in terms of the correct decision rate. As a case study, it is applied to a particular face recognition classification algorithm based on linear discrimination analysis. The performance comparison with a traditional approach has proven the value of the proposal.

Face based recognition algorithms: The use of uncertainty in the classification

The paper proposes a new approach to a decision making process based on face recognition algorithms. The proposed approach is based on the use of the uncertainty on the information used in the classification process to provide suitable level of confidence to the system output. As a case study, it is applied to a specific face recognition classification algorithm, namely the one based on Linear Discrimination Analysis, but it can be easily extended to other classification algorithms even dealing with other applications.

Performance Evaluation of Face Classification Systems

In this paper face classification systems based on 3D images are compared in terms of classification and metrological performance in presence of image uncertainty. In previous papers the authors proposed a new approach to classification and recognition problems. It is based on the evaluation of the image uncertainty and on the exploitation of such information to provide the confidence level of classification results. Such approach is here adopted for comparing several 3D architectures, different for camera specifications and geometrical positioning, with the aims of quantifying their performance from a metrological point of view and of identifying the configuration able to optimize the result reliability.

A proposal for improving the performance of face recognition systems based on 3d features

In this paper a suitable methodology for the improvement of the reliability of results in classification systems based on 3D images is proposed. More in detail, it is based on the knowledge of the uncertainty of the features constituting the 3D image (obtained processing a pair of two 2D stereoscopic images) and on a suitable statistical approach providing a confidence level to the classification result. These pieces of information are then managed in order to improve the classification performance in terms of correct classification and missed classification percentages. The experimental results, obtained applying the methodology on an Active Appearance Models algorithm, a popular method for face recognition based on 3D features, show that, compared with a traditional approach (which generally does not take into account the uncertainty on 3D features), the proposed methodology allows to significantly improve the classification performance even in scenarios characterized by a high uncertainty.

The use of uncertainty for improving the reliability of classification in face based recognition algorithms

The paper proposes a new approach to classification and recognition problems which takes into account the measurement uncertainty affecting input data for improving the overall reliability of such kind of processes. The proposed method is based on an effective probabilistic approach for the evaluation of the confidence level of system outputs and the suitable use of such information for improving the performance in terms of correct decision.

Influence Quantity Estimation in Face Recognition Digital Processing Algorithms

Nowadays, the matter of uncertainty in face recognition-based biometric systems is a relevant issue for the scientific community. This is due to the even more increasing deployment of such systems in critical applications such as safety, security, and access control, to cite a few. In this context, the authors are engaged in the design of general methods for uncertainty modeling and evaluation aimed at realizing face recognition-based biometric systems with built-in uncertainty evaluation capability. In this way, the output of a recognition system will not be the identity of the observed subject, but a confidence level for each possible subject. In previous papers the authors have identified the quantities of influence and have proposed a suitable uncertainty model. The core of the proposed model is the knowledge of the value assumed by the quantities of influence with respect to the corresponding values achieved in suitable reference conditions. This paper mainly analyzes these measurement issues, a fundamental step toward the development of such systems with built-in uncertainty evaluation capability. First results show a good agreement between statistical indicators and a priori estimations achieved with the proposed method.