Enhanced Offline Signature Verification using GLCM and Correlation-based Feature Selection Technique

In security sensitive applications, signature verification plays vital role in identity confirmation. Th feature extraction and feature selection process is crucial for accomplishing the task of authentication. In this paper offline signature verification system is proposed, which mainly concentrates on extracting proper features from signature images. Then irrelevant and redundant features are removed using a feature selection method. Gray-Level Co-occurrence Matrix (GLCM) is employed as a texture feature extractor, followed by correlation-based feature selection (CBS) to refine the feature space and enhance classifier performance. The goal is to find the most discriminative features required for distinguishing authentic and forged signatures. The outcome of the proposed method is the optimized set of features that are fed into an ensemble voting classifier comprising a Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Random Forest (RF). The proposed approach is tested on four different publicly available benchmark datasets, including CEDAR, BHSig260-Bengali, BHSig260-Hindi, and MCYT. These datasets comprise diverse signature styles and demographic variations, which are required for developing a more generalized model. Results demonstrate the superiority of the proposed methodology, achieving reduced feature dimensionality resulting in low computational complexity without compromising decision consistency. This paper highlights the potential of combining optimized texture-based feature extraction with ensemble learning for efficient and accurate offline signature verification, facilitating practical applications in resource-constrained devices.