Siganture verification using a multidimensional Gaussian distribution

OData support
Dr. Kővári Bence András
Department of Automation and Applied Informatics

During the semester I created a classification algorithm that uses multivariate normal distributions to create models that approximate data extracted from signatures. Data are grouped based on which feature they belong to within a signature, because I made the assumption that there is a strong correlation between these values, so it is worthwhile to investigate them together. After building a model, I check whether the investigated signatures fit the model, and make a decision based on that whether they are to be considered original or forged. I then integrated the created algorithm into an existing system by implementing the necessary interfaces. I also added new controls to the existing user interface so it would be possible to control parameters of the multivariate normal distributions. I also tested the created algorithm on a given set of signatures and investigated its efficiency and usability. After evaluation, I investigated where the algorithm could be used and in what direction it should be developed further. To ease the processing of data I created filters that can be configured with each input set and afterward be reused in both the training and testing phases.


Please sign in to download the files of this thesis.