Offline Signature Verification Model Using CNN and PSO Algorithm

The handwritten signature verification plays a crucial role in various applications including authentication and document verification. The feature extraction stage of the offline signature verification system is considered essential and significantly affects the performance of the system. The quality and quantity of the extracted features play a vital role in the ability of the system to differentiate between genuine and forged signatures. In this study, we propose a novel approach for optimizing the hyperparameters of a Convolutional Neural Network (CNN) model for handwritten signature verification using the Particle Swarm Optimization (PSO) algorithm. The PSO algorithm is a population-based optimization method that draws inspiration from flocking behavior in birds. We defined a search space consisting of hyperparameter ranges, including the number of convolutional filters, filter sizes, number of dense layers, dropout rate, and learning rate. The PSO algorithm explores this search space by iteratively updating the positions and velocities of the particles to determine the best set of hyperparameters that maximizes the accuracy of the CNN model. We evaluated our approach on the BHSig260-Bengali, BHSig260-Hindiin, GPDS, and CEDAR datasets, which contain a diverse collection of handwritten signature images. The results of the experiment demonstrated that the proposed approach obtained a high accuracy of 98.3% on the testing dataset.