Context-Aware Palmprint Recognition via a Relative Similarity Metric

TL;DR

The paper addresses the limitation of traditional palmprint matching, which relies on isolated pairwise similarity, by introducing a Relative Similarity Metric ($R$) that captures contextual consistency of similarities across the gallery. RSM is defined as $R(x_i, x_j) = \frac{1}{M} \sum_{x_k \in \mathcal{D'} \setminus \{x_i, x_j\}} \Delta_{i,k}^{(j)} + \alpha S(x_i, x_j)$ with $\Delta_{i,j}^{(k)} = |S(x_i, x_k) - S(x_k, x_j)|$, using $M=30$ and $\alpha=1$, and is applied on top of the CCNet hashing/classification pipeline. This relational augmentation yields a new state-of-the-art Equal Error Rate (EER) of $0.000036\%$ on the Tongji palmprint dataset, improving over the CCNet baseline of $0.000042\%$ by better suppressing impostor-like high similarities and genuine misrankings. The findings demonstrate that incorporating dataset-wide relational structure into biometric matching enhances discriminability and robustness, with potential applicability to other modalities and large-scale retrieval tasks.

Abstract

We propose a new approach to matching mechanism for palmprint recognition by introducing a Relative Similarity Metric (RSM) that enhances the robustness and discriminability of existing matching frameworks. While conventional systems rely on direct pairwise similarity measures, such as cosine or Euclidean distances, these metrics fail to capture how a pairwise similarity compares within the context of the entire dataset. Our method addresses this by evaluating the relative consistency of similarity scores across up to all identities, allowing for better suppression of false positives and negatives. Applied atop the CCNet architecture, our method achieves a new state-of-the-art 0.000036% Equal Error Rate (EER) on the Tongji dataset, outperforming previous methods and demonstrating the efficacy of incorporating relational structure into the palmprint matching process.

Figures (2)

• Figure 1: Comparison of score distributions for match and non-match pairs using cosine similarity versus the proposed Relative Similarity Metric (RSM).
• Figure 2: Comparison of a subset of 'difficult' score distributions for match and non-match pairs using cosine similarity versus the proposed Relative Similarity Metric (RSM).