Handwriting biometrics applications in e-Security and e-Health are addressed in the course of the conducted research. An automated graphomotor analysis method for the dynamic electronic representation of the handwritten signature authentication was researched. The developed algorithms are based on dynamic analysis of electronically handwritten signatures employing neural networks. The signatures were acquired with the use of the developed electronic pen described in the paper. The triplet loss method was used to train a neural network suitable for writer-invariant signature verification. The signatures represented by anchor and positive are master signatures from the same class, representing a single person, that should be close to each other in multidimensional output space. For each signature of the three, the same neural network calculates a fixed-length latent space representation. The hand-corrected dataset containing 10,622 was used in order to train and evaluate proposed network. After learning, the network was tested and achieved mean 5.77% EER. The use of the triplet loss algorithm to teach neural network generation of embeddings has proven to give good results in terms of grouping of similar signatures and separating them from signatures representing different people.
Authors
Additional information
- DOI
- Digital Object Identifier link open in new tab 10.3390/electronics10040456
- Category
- Publikacja w czasopiśmie
- Type
- artykuły w czasopismach
- Language
- angielski
- Publication year
- 2021
