Review of Signature Recognition Using Deep Learning with Dataset Profiling for Career Pathway Recommendation

Abstract

While graphology has historically faced scepticism due to subjectivity, recent Deep Learning (DL) integrations—specifically Convolutional Neural Networks (CNN)—offer a path toward objective personality assessment. However, a significant research gap persists in the systematic mapping of DL-extracted signature features to standardized psychological frameworks for career guidance. This review synthesizes current literature at the intersection of signature-based graphology, the Holland RIASEC model, and CNN architectures. We critique existing feature extraction techniques and classification frameworks, revealing a critical lack of large-scale, standardized datasets and a "black-box" interpretability problem that hinders clinical adoption. By evaluating the synergy between automated trait detection and career typology, this paper identifies three primary research voids: the absence of cross-cultural validation, the need for hybrid models that combine psychological theory with raw pixel data, and the ethical implications of AI-driven recruitment. Ultimately, this review provides a roadmap for transitioning AI-enhanced graphology from a niche experimental tool to an evidence-based framework for personalized career counselling.