Presentation of Master's Thesis

Estudio de la biometría del trazo aéreo de la firma para la identificación personal

Images

ilustracion 1 smallilustracion 2 smallDate: junio de 2015
Author: Elyoenai Guerra Segura
Tutors: Dr. Carlos Manuel Travieso González and Dr. Jesús Bernardino Alonso Hernández

 

Abstract

This work presents an initial study about the biometry of the in-air signature for the personal identification.

First of all the Leap Motion sensor is evaluated to be used in order to model aerial signatures. Secondly, the temporal variability of the different generated signals is studied.

Next, Support Vector Machines classifier is applied with and without a previous feature selection step. The feature selection step is based on the well-known technique of Principal Component Analysis. Moreover, an automatic classifier based on the Dynamic Time Warping algorithm is also studied.

According to the results, Leap Motion offers a wide variety of signals generated from the user’s aerial path. These signals are evaluated during 4 months for the same user in order to study the intra-user temporal variability. Results show that some of these signals are more stable than others, what means they are more independent from the moment the signature is made.

The achieved results are better than the ones reported by the known previous works in terms of Equal Error Rate. The generated database was formed of 39 users with 10 samples per each user. SVM offers an EER of 0.36%, while DTW gives an EER of 0.51%.

The results are promising since this work is an initial study of the problem. Based on this work, a possibility to develop more specialized researches is created.