Z-ENG: Online signature verification using Machine-learning algorithms

2024-2025 tavasz

Szoftver

Téma leírása

Signatures are one of the most commonly used biometrics for user recognition or authentication. Based on the signature input method, it is categorized as online or offline. Offline are the regular signatures acquired using ink pen and a paper then scanned as a digital file. While in online signatures, special devices like tablets, which have the ability to capture several features like position, pressure, altitude, and azimuth, are used to acquire the signatures.

The online signature data file contains all the points of the signatures with its all-available features. A point in a digital signature might look like this “ 55 102 5455699 0 145 “, where each number represents a feature (eg: xy pressure pen-up altitude).

To verify if a signature is genuine or forged many verification methods can be used to check the similarity between the tested signature and the reference signatures. In this topic, we want to test some Machine-Learning approaches for this purpose. We will use part of the signature database to train the system, then test it with the rest of the data and try to achieve the maximum accuracy.

This is part of a bigger signature verification project, where many verification approaches, preprocessing methods, and loaders are already implemented and can be used for the initial steps.

The preferred candidate for this topic would be someone with good knowledge of machine learning algorithm and good programming skills in Python or C#.

If you feel that this topic suits you, please send me a description about yourself and your skills, and a brief description or proposal of the work idea you want to apply regarding this topic if you have one.

Steps towards completing a task on this topic:

- Read and increase your knowledge about online signature verification

- Read and increase your knowledge about machine learning

- Apply the first few steps of a verification system using python or C# (loading data, data preprocessing and enhancement, feature extraction, ..)

- Choosing the machine-learning algorithm/s you want to use and the verification approach, and then apply it to your work

- Study and analyze the results of your verification system, increase the accuracy if possible, compare the results and conclude your work.

- Write your thesis (thesis level) / prepare a presentation (project level)

Feltételek

  • machine learning

Maximális létszám: 3 fő