1. A mechanism for verifying signatures using multi-staged approaches involving preprocessing, segmentation, and feature extraction. 2. An approach that involves utilizing DBLSTM, DTW, and comparators to enable accurate signature matching. 3. A score fusion unit which helps to aggregate various similarities in order to enhance the performance of signature matching. 4. A decision making unit that uses the similarity scores in determining whether the signature is a forgery or genuine. 5. A computer readable medium comprising executable instructions for verifying and classifying signatures through comparisons.

Description:[0003] Various contemporary uses of handwritten or behavior-based signatures require efficient methods of verifying their legitimacy for proper authentication purposes to prevent any form of unauthorized access or fraud. The classification approach based on signatures finds application in domains such as computer and information security surveillance, fraud identification, network traffic examination, identification, and intelligence gathering operations where records or signatures to be verified can be matched to certain reference signatures.
[0004] Signature verification systems typically use a comparison between a signature being analyzed and one or several reference signatures to ascertain whether it is genuine or forged. This type of verification may apply to a variety of uses, including financial transaction verification, entry into restricted areas, documents validation, or digital identities management systems.
[0005] Single comparison approaches, as well as single-classifier-based solutions for signature verification are known to have some shortcomings as they do not necessarily consider all aspects of signatures that may include temporal, spatial and dynamic features. Moreover, known systems may demonstrate reduced accuracy when facing intra-class variability and well-trained forgeries, as well as differences associated with varying acquisition equipment and conditions.
5
[0006] In addition, any system that uses single decision-making process is vulnerable to the influence of noise or errors of segmentation or extraction of features, leading to false classification results. Therefore, the present invention aims at development of more advanced technology that will incorporate multiple methods of analysis, as well as classification and fusion of results produced by multiple classifiers. , Claims:1. A mechanism for verifying signatures using multi-staged approaches involving preprocessing, segmentation, and feature extraction. 2. An approach that involves utilizing DBLSTM, DTW, and comparators to enable accurate signature matching. 3. A score fusion unit which helps to aggregate various similarities in order to enhance the performance of signature matching. 4. A decision making unit that uses the similarity scores in determining whether the signature is a forgery or genuine. 5. A computer readable medium comprising executable instructions for verifying and classifying signatures through comparisons.