Original title: Video-Based Palm-Vein Authentication under Challenging Conditions
Xiaofeng Yan, Kechen Liu, Abhilash Venkatesh, Cathy Zhang +2 · September 2026
Scanning the vein patterns inside your palm is a popular way to confirm your identity without touching a screen. But in daily life, hands get sweaty, dirty, or warm, which can hide those veins. Because researchers lacked realistic data, they built the Columbia University Palm-vein dataset, recording palm videos under clean, warm, wet, and dirty conditions. They found that standard systems lose most of their accuracy on dirty palms, with error rates roughly quadrupling. To solve this, the researchers created a two-part method. Over time, it combines information across video frames to cancel out temporary blurs. Across space, a test-time matcher uses saliency to steer comparisons around dirty patches, routing around corrupted regions. This approach achieves top performance across every tested surface condition while remaining small and fast. It also improves existing systems by 29-37% without needing to retrain them.
Imagine you're trying to recognize a friend's handwriting on a letter that got smudged in the rain. Parts are blurry or obscured, so you can't rely on the whole page at once. Instead, you glance at several different angles where the ink might be clearer, and you focus on the strokes that aren't smudged. In this paper, the palm-vein scanner faces similar issues with dirty or wet hands, and the method uses multiple video frames and concentrates on uncorrupted regions to make a better guess.
Palm-vein authentication is used for secure access, but real-world conditions like sweat or dirt can make it unreliable. This work provides a way to maintain accuracy without retraining systems, making biometrics more practical in everyday settings.
The abstract notes that the audit found demographic gaps in warm conditions, specifically related to body water and gender, that persist even after correction. The dataset is for non-commercial research only, and the method's performance on other types of biometrics or in even more extreme conditions isn't addressed.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Video-Based Palm-Vein Authentication under Challenging Conditions — Xiaofeng Yan, Kechen Liu, Abhilash Venkatesh et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.