SPECTRUM: Empowering Online Handwriting Verification via Temporal-Frequency Multimodal Representation Learning

Peirong Zhang, Lianwen Jin

Keywords: Online handwriting verification; Multimodal representation learning; Temporal and frequency learning; Handwritten biometrics

TL;DR: We propose a multimodal temporal-frequency synergistic model to empower online handwriting verification.

Abstract: Tapping into the uncharted multimodal representation learning in online handwriting verification (OHV), we propose SPECTRUM, a temporal-frequency synergistic model tailored to enhance handwriting representations. SPECTRUM comprises three core components: (1) a multi-scale interactor that interweaves fine-grained temporal and frequency features across multiple scales through complementary domain interaction; (2) a self-gated fusion module, dynamically integrating global temporal and frequency features via self-driven balancing. Collectively, these two components achieve micro-to-macro multimodal integration; (3) a multimodal distance-based verifier that fully harnesses temporal and frequency representations, sharpening genuine-forged discrimination beyond conventional temporal-only approaches. Extensive experiments demonstrate SPECTRUM's pronounced outperformance over existing OHV methods. Furthermore, we reveal that incorporating multiple handwritten biometrics fundamentally improves the discriminatory power of individual writing features. These findings not only validate the efficacy of multimodal learning in OHV but also encourage broader multimodal research across both feature and biometric domains, potentially opening new avenues for future explorations. Code will be publicly available.

Primary Area: unsupervised, self-supervised, semi-supervised, and supervised representation learning

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 10403