DWSCNN Online Signature Verification Algorithm

Signature verification has always been a hot topic in biometrics field. Recent advancements in signature verification have been greatly influenced by deep learning technologies, particularly deep neural networks. While these advancements have notably boosted the accuracy of signature verification systems, practical implementation still faces hurdles.Our paper introduces a novel algorithm tailored for online handwritten signature verification. Our method leverages a Depth-wise Separable Convolutional Neural Network (DWSCNN) that relies on Depth-wise separable convolution for analyzing and confirming signature feature sets. In comparison to conventional Convolutional Neural Networks (CNNs), the DWSCNN significantly reduces the number of neural network parameters while maintaining a similar level of classification performance. This reduction translates to considerably shorter processing times and lower resource requirements. On average, our algorithm achieves an impressive verification accuracy of 96.92%, outperforming mainstream online signature verification frameworks. This underscores the efficiency and efficacy of our proposed framework.