Interval-valued writer-dependent global features for off-line signature verification

Manjunatha, K. S. and Guru, D. S. and Annapurna, H. (2017) Interval-valued writer-dependent global features for off-line signature verification. In: 5th International Conference on Mining Intelligence and Knowledge Exploration  MIKE 2017, December 13–15, 2017, Hyderabad.

Full text not available from this repository. (Request a copy)


This work focuses on the proposal of a method for Off-line signature verification based on selecting writer-dependent global Features. 150 Global features of different categories namely geometrical, texture based, statistical and grid features for offline signatures are computed. Writer dependent features are selected through an application of a filter based feature selection method. Further, to preserve the intra-writer variations effectively, the selected features are represented by interval-valued data through aggregation of samples of each writer. Here in this work, we recommend creating two interval valued feature vectors for each writer. Decision on the test signature is accomplished by means of a symbolic classifier. In the first stage, we conducted experiments with writer dependent features by keeping a common dimension for all writers. Further, we conducted experiments with varying writer dependent feature dimension and threshold as done by a human expert. To demonstrate the effectiveness of the proposed approach extensive experimentation has been conducted on both CEDAR and MCYT offline signature datasets. The Error-rate obtained with the proposed model is low in comparision with many of contemporary models.

Item Type: Conference or Workshop Item (Paper)
Subjects: D Physical Science > Computer Science
Divisions: Department of > Computer Science
Depositing User: C Swapna Library Assistant
Date Deposited: 07 Mar 2020 05:24
Last Modified: 11 Mar 2020 09:00

Actions (login required)

View Item View Item