Recent Advances in Offline Signature Identification

Abstract

In recent years, biometric-based authentication systems have been widely used in many applications which require reliable identification scheme. Among others, handwritten signature is one of the most interesting biometric means, that is being considered with renewed interest. This paper presents some of the most relevant advances in the field of offline signature identification and highlights some directions for further research.In recent years, biometric-based authentication systems have been widely used in many applications which require reliable identification scheme. Among others, handwritten signature is one of the most interesting biometric means, that is being considered with renewed interest. This paper presents some of the most relevant advances in the field of offline signature identification and highlights some directions for further research

An Efficient Signature Verification Method Based on an Interval Symbolic Representation and a Fuzzy Similarity Measure

“…In the off-line handwritten signature verification area, a significant volume of research has been undertaken for the development of different authentication methods [1-14, 19, 23-45, 51-54]. Reviews of different methods proposed for signature identification/verification at different periods of time can be found in [1,[23][24][25][26]. Recently, some competitions have also been organized to evaluate the technological achievements in terms of reliability and accuracy in the field of signature verification [2,9].…”

“…For the writerindependent approach, on the other hand, a single model is trained for all the signature classes to be used for the verification task. Hybrid models have also been used for signature verification [1,[23][24][25][26].…”

“…From the literature reviewed, it is noted that there has been significant progress in the off-line signature verification domain. However, despite the progress in the area over the past decades, it remains an open research problem [1,2,23,24]. In addition to the limitations mentioned in Table 1, some general challenges [24] in the area of off-line signature verification that still attract many researchers for further investigation in this field are indicated as follows: i) high intra-class variability in handwritten signatures of every individual compared to the physiological biometrics, such as fingerprints or iris of the individual, ii) low inter-class variability between genuine signatures and skilled forgeries of every individual, iii) the existence of only genuine signatures as partial knowledge for training off-line signature verification systems, iv) limitations in the amount of signature data available for training off-line signature verification systems in real scenarios, as during the enrolment process users often provide only a few samples of their signatures, and v) the presence of signatures written in different scripts.…”

“…Signatures, as one of the behavioural human characteristics, are extensively used as a proof of identity for legal purposes on many documents such as bank cheques, credit cards, and wills in our daily lives. Considering the large number of signatures handled daily through visual inspection by authorized persons, construction of an efficient automatic system to handle such a huge volume of signatures has many potential benefits for signature authentication to reduce fraud and other crimes [1,[23][24][25][26].…”

“…A quick look at the literature of signature identification / verification indicates that handwritten signature identification / verification systems are well established. A wide range of algorithms have already been developed in the past few decades to automatically process handwritten signatures in various signature-based applications, such as person identification / verification, cheque fraud detection, bank transactions, and crime detection [1,[23][24][25][26]. Considering the way in which the proposed methods in the literature dealt with the handwritten signatures, the methods can be categorized into two groups: a) identification, and b) verification [4].…”