An Offline Tool to Verify Signature for Indian Languages in Digital Forensics

Signature verification is a crucial aspect of digital forensics as it helps identify an individual's identity when validating various documents such as bank cheques. Since signatures are language-dependent, different states in India have different languages, making it essential to consider native languages and combinations of signatures for research purposes. However, publicly available datasets for Kannada, Hindi, and English language signatures are limited. Consequently, the creation of a database and techniques to identify signatures in these languages becomes necessary. This paper employs a series of processes including pre-processing (binarization, noise removal, and thinning), feature extraction (global and grid-based features), and verification of test signatures using Self Organized Maps (SOM). The system yields verification results for combinations of Kannada, English, and Hindi languages. With an overall identification rate of 94.3% accuracy, the system demonstrates successful signature detection across all three languages. Moreover, individual language accuracies for Kannada, English, and Hindi are at 99.5%, 84.5%, and 98.5% respectively. Additionally, the system's performance can be further enhanced by employing different classification and feature extraction techniques.