Research
Research focuses on two main areas:
1) Image Processing and Biometric Systems
2) Human-Computer Interaction, with a focus on design and usability evaluation of interactive systems and e-learning, and design for users with special needs
The meeting point between the two areas is identified in the role of the image in the interaction between man and machine in many contexts.
The interest in image processing has spurred activities centered on Biometric Systems. Currently, these systems are of great importance in security and user identification both local (protected areas) and remote (e-services). Particularly promising are systems based on face recognition, and those based on the iris, as minimally invasive and pretty accurate. However, both suffer from problems and limitations that must be overcome by research. For the face, particular attention should be paid to changes in pose, illumination, expression (PIE), and to occlusions. For some of these problems, in particular the occlusions, fractals used locally have proved an appropriate solution. A more recent line of research deals with face recognition after plastic surgery. The face recognition in non-optimal conditions is also used for indexing of video sequences according to the present subjects, which can have both applications related to video surveillance, and to the cataloging or recognition of statistics of various type (for example the duration of intervention during a debate). For the iris, problems to be solved are mostly due to the acquisition at a distance and to illumination. In particular, the BIPLab group at University of Salerno, whit which we work in this context, has achieved the sixth position in the international competition NICE II, who had precisely the aim of comparing different techniques for iris recognition in non-optimal conditions. One alternative biometrics is ear, that the work of Iannarelli have proven to be an excellent mean of identification. In any case, the current trend is to use multibiometric systems, in which each module has the potential to compensate for the limitations of the other. In such settings, it is necessary to normalize the values of the scores assigned during comparisons before arriving at the final result. Moreover the different classifiers should be combined on a flexible basis, and possibly reconfigured over time. An important role is attributed to the measures of the quality of the biometric sample and reliability of the responses, especially in the case of real applications in non-controlled settings. We provided innovative solutions for both problems. Systems of this type can also be used in the context of the creation of intelligent environments, i.e. capable of reacting in different ways depending on the identity of the user who uses them. We also studied strategies for dealing with spoofing in the case of authentication through video of the face. The last line of research about the biometrics concerns the calculation of the entropy of a gallery of registered subjects, considered in terms of representativeness of the templates stored in order to facilitate recognition in different contexts (different devices) and environmental (PIE - Pose , Illumination, Expression).
Some publications in this area:
For users with special needs, distance learning can be an important occasion for the use of services otherwise often denied. For this reason, we have presented a proposal for an alternative strategy for the design of educational materials, which includes the involvement in the early stages of disabled users and users with experience in teaching the disabled. This strategy should lead to the implementation of systems with appropriate translation modules and interfaces, to enable interaction in different alternative ways. In particular, we are studying the mode of interaction and presentation of the material for deaf students. We are working to design and build tools to enable the creation and management of documents in SignWriting, a written form of sign language that seems to be particularly appreciated by deaf users. For automatic analysis of video in sign language we have also begun a collaboration with Professor Patrice Dalle from the University Paul Sabatier of Toulouse, which includes the activity of co-tutoring a PHD student.
Still in the field of computer-aided learning, we are experiencing the integration of the concept of Zone of Proximal Development devised by the Soviet psychologist Lev S. Vygotsky in the architecture of a system for the creation and customization of didactic and collaborative activities.
