Inspecting Offline Handwritten Signature Intra-Variation Over Time: An Empirical Study

Handwritten signatures serve as crucial personal identifiers and have been extensively used for authentication purposes for a long time in the human race. Signatures exhibit substantial variations influenced by factors such as mood, time, writing speed, and the writing tool used. Understanding the variations that occur in offline handwritten signatures over time for an individual is paramount in forensic investigations, biometric systems, and legal document authentication, even in this digital era. This article presents an empirical study focused on inspecting the intra-variation of offline handwritten signatures over an extended period. To conduct this study comprehensively, we collected an extensive dataset comprising 6400 signature samples from 100 distinct writers scribbled intermittently over several months, providing a rich and diverse set of signatures for analysis. To analyze the collected data effectively, we ensembled some contemporary convolutional models to train a deep architecture. Our experimental results were quite encouraging and may shed light on the tendencies and patterns in signature changes by providing valuable insights for biometric system design and forensic experts.