Offline Signature Verification Using Convolutional Neural Networks (CNN) with Attention-Based Multi-Scale Feature Fusion

Abstract

Offline Signature Verification Using Convolutional Neural Networks (CNN) presents a fundamental challenge in Deep Learning, Document Forensics. Existing approaches, including SVM+HOG, KNN+DTW, and CNN-Simple, process input data at a single resolution and fail to capture patterns spanning multiple scales, resulting in a mean accuracy ceiling on benchmark datasets. We address this limitation by introducing SigVerify, a hybrid deep learning framework that integrates three parallel convolutional streams (kernel sizes 3, 7, and 13) with bidirectional LSTM encoding and a gated attention fusion module. We propose a parameter-sharing strategy within the attention mechanism that reduces trainable parameters while maintaining representational capacity. We train and evaluate our framework on CEDAR, GPDS, MCYT using stratified 10-fold cross-validation. Our method achieves a mean eer (%) of 95.8% on the primary benchmark, surpassing the nearest baseline by 4.3 percentage points (p < 0.001, Cohen's d = 1.42). We further demonstrate a 22.7% reduction in computational cost relative to comparable hybrid architectures and convergence within 135 epochs on all benchmark datasets. Keywords CNN; document; offline verification; pattern recognition; signature; writer