Publication:
Evaluación de usabilidad de un sistema de reconocimiento biométrico multimodal accesible

Loading...
Thumbnail Image
Identifiers
Publication date
2014-06-22
Defense date
2014-07-15
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Google Scholar
Export
Research Projects
Organizational Units
Journal Issue
Abstract
El avance de las tecnologías de la información supone un uso cada vez mayor de los sistemas de reconocimiento biométrico, tales como la firma manuscrita o la huella dactilar para tareas cotidianas. Ambas modalidades biométricas aportan mayor comodidad, reducen el tiempo necesario para llevar a cabo una actividad y disminuyen el esfuerzo requerido para ello; para la mayoría de la sociedad supone una ventaja a la hora de realizar ciertos trámites, sin embargo, al colectivo de tercera edad le supone una barrera en muchos casos ya que es un entorno desconocido para ellos en la mayoría de los casos El objetivo de este estudio es analizar el comportamiento de este colectivo al someterse a la verificación tanto de la firma manuscrita como de la huella dactilar, para ello se han realizado análisis de usabilidad y accesibilidad, así como de la evolución del rendimiento de los algoritmos biométricos.
We are currently experiencing a revolution of information in which new technologies have great impact. These changes make it necessary to acquire knowledge continuously. Technology has an important role in many areas and it has helped to progress in different fields out of technology like medicine for example. Therefore, the society must adapt to these changes. This revolution involves a great advantage for many companies even increasing the easiness of the final products in many cases. However, for some people instead of being an advantage that is a disadvantage. This is because the advancement of technology is faster than the learning. This phenomenon is called "digital divide". Every day there are more areas in which new technologies are used and this is a problem for that people who did not integrate the technology into their daily lives. This project has been done to gather information about the usability and accessibility of certain biometric systems. This work covers a gap in the state of the art because there are only a few studies on this with this sector of the population. For this experiment we have selected a total of 28 elder people with different gender, different level of education, different knowledge of the technology and different jobs. We used two different biometric modalities: handwritten signature and fingerprint. The main difference between these two is that the handwritten signature is a behavioural modality and fingerprint is physiological. The fingerprint identification is safer because it is a personal, unique and transferable property; but its use is less widespread than the handwritten signature. To perform the analysis first an application meeting special requirements (e.g. easy and intuitive) was designed. Then, the evaluation process consists of 3 parts that have been split on two visits to each user. The first two parts are covered in the first session: the recruitment and data collection process. In the second visit the data collection process is repeated. The two views are separated in time between 2 and 5 days for users non-familiarized with technologies and between 10 and 15 days for those who do know them. In the personal data recruiting process users were requested for their age, name, gender, accessibility issues, laterality and previous jobs. Three accessibility variables were also selected by users at the recruiting process: the screen brightness, volume and colour of the app theme. The second process is the training test of both biometric modalities. Then, users proceed with the enrolment phase of their biometric traits: fingerprints of the index and ring fingers of both hands and a handwritten signature are stored in the database. Then, the app guides the user through two steps more: a simulated purchase using NFC (where the user authenticate himself with the fingerprint or the signature) and a multiple fingerprints and signatures collection in order to build a database. The purchase process is made bringing the mobile device to the NFC card, selecting the mode in which you want to verify the purchase, identifying with this mode (there are 3 attempts) and bring the device back to the NFC card to confirm your purchase. The second part (the process of data collection) is done by first capturing fingerprints and secondly capturing handwritten signatures. For each registered finger five samples are acquired, with 3 attempts for obtaining a correct fingerprint in each catch. For the handwritten signature 10 samples are captured with 3 trials for each of them. The third part takes place on the second visit and the process is the same as in the second part. This is useful to evaluate learning and efficiency. The captured data is confidential so it must be protected. Before starting the experiment the user is asked to sign a paper that meets the National Data Protection Law. After the second visit users fill out a satisfaction form anonymously to carry on later usability analyses of the data. With the gathered data an analysis of usability, accessibility and performance has been done. For the analysis of usability three variables have been taken into account: effectiveness, efficiency and satisfaction. The number of users who have completed the process correctly (effectiveness) is 74%. Regarding the purchases, 95.95% of them have been completed successfully. Efficiency is the time spent in a task: 91% of users have reduced the time on the second visit compared to the first. Users evaluated the system globally satisfactory and 100% of them would use similar identification systems in the future. Regarding the accessibility parameters, 96.43% of the users have chosen medium volume and the rest high volume; 50% of them have selected medium brightness, 46% high brightness and the rest low brightness; regarding the colour, 46.42% of the users have selected white colour, 39.28% black and the rest blue. The performance analysis has been done using two algorithms (for fingerprint and signature recognition). They were used during the evaluation and after it in order to process the data. For the performance calculation, the users have been divided into two groups: those that have previous knowledge about technologies and those that haven't. The performance during the evaluation in terms of genuine rates was 7.36% for users who have familiarity with the technology and 6.65% for users who have not previous experiences with the technology. The fingerprint error rates that are much higher, are very close in both groups and for both visits (33.7% for users familiarized with the technologies and 31.41% for users that are unfamiliar). For the posterior performance processing three types of common error rates (FAR, FRR, EER) were used. The FRR is the result of comparing samples from the same user (genuine results), the FAR is the result of comparing samples from one user with samples another users randomly (zero effort impostor users). The EER is the rate at which both acceptance (FAR) and rejection (FRR) errors are equal. The final results are given with the ROCs below (left-Handwritten signature / right - Fingerprint). The main conclusion is that identification via fingerprint is not suitable for older people because the error rates are very high, so in the future, other biometric identification methods such as iris detection should be investigate.
Description
Keywords
Biometría, Usabilidad, Algoritmos, Radiocomunicación
Bibliographic citation