Optimization approaches in DTW based on-line siganature verification

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

Automatic signature verification is a hot topic nowadays. Most documents must be signed at least by one person in order to validate them. Manual verification of signatures is a time-consuming task, so it is important to replace this process with a computer program.

In my dissertation I will present the signature verification program that I created which performs the classification of the signatures using the Dynamic Time Warping (DTW) algorithm. I have also created some data cleaning techniques that will help the algorithm work efficiently. I constructed an overall statistical table to help determine the effectiveness of the algorithm and the individual data cleaning techniques.

On average, the DTW algorithm leads to an average of about 75% of the cases in terms of the authenticity of the signatures. By using the data cleaning techniques correctly, this result can be increased by up to 10%.

The program I have created shows the efficiency of the algorithm used and the increase of it with the help of the implemented data modification methods.

Downloads

Please sign in to download the files of this thesis.