Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

To Impute or Not: Recommendations for Multibiometric Fusion

Melissa R Dale, Elliot Singer, Bengt J. Borgström, Arun Ross

TL;DR

This work tackles the challenge of missing match scores in multibiometric score-level fusion by systematically evaluating both univariate and multivariate imputation strategies across three diverse multimodal datasets. It employs simulated missingness up to 90% and analyzes the impact of training data balance and inter-modality correlations on imputation effectiveness, with MICE-based multivariate methods—especially Bayesian Ridge regression—showing robust gains for missing data. The findings emphasize that imputation outperforms simply discarding incomplete vectors, that balancing the training set mitigates biases toward overrepresented classes, and that the choice between multivariate and univariate imputation should reflect inter-modality score correlations. The results offer practical guidance for deploying imputation in real-world biometric fusion systems and point to avenues for future work in label-free, hybrid, and context-aware imputation strategies.

Abstract

Combining match scores from different biometric systems via fusion is a well-established approach to improving recognition accuracy. However, missing scores can degrade performance as well as limit the possible fusion techniques that can be applied. Imputation is a promising technique in multibiometric systems for replacing missing data. In this paper, we evaluate various score imputation approaches on three multimodal biometric score datasets, viz. NIST BSSR1, BIOCOP2008, and MIT LL Trimodal, and investigate the factors which might influence the effectiveness of imputation. Our studies reveal three key observations: (1) Imputation is preferable over not imputing missing scores, even when the fusion rule does not require complete score data. (2) Balancing the classes in the training data is crucial to mitigate negative biases in the imputation technique towards the under-represented class, even if it involves dropping a substantial number of score vectors. (3) Multivariate imputation approaches seem to be beneficial when scores between modalities are correlated, while univariate approaches seem to benefit scenarios where scores between modalities are less correlated.

To Impute or Not: Recommendations for Multibiometric Fusion

TL;DR

This work tackles the challenge of missing match scores in multibiometric score-level fusion by systematically evaluating both univariate and multivariate imputation strategies across three diverse multimodal datasets. It employs simulated missingness up to 90% and analyzes the impact of training data balance and inter-modality correlations on imputation effectiveness, with MICE-based multivariate methods—especially Bayesian Ridge regression—showing robust gains for missing data. The findings emphasize that imputation outperforms simply discarding incomplete vectors, that balancing the training set mitigates biases toward overrepresented classes, and that the choice between multivariate and univariate imputation should reflect inter-modality score correlations. The results offer practical guidance for deploying imputation in real-world biometric fusion systems and point to avenues for future work in label-free, hybrid, and context-aware imputation strategies.

Abstract

Combining match scores from different biometric systems via fusion is a well-established approach to improving recognition accuracy. However, missing scores can degrade performance as well as limit the possible fusion techniques that can be applied. Imputation is a promising technique in multibiometric systems for replacing missing data. In this paper, we evaluate various score imputation approaches on three multimodal biometric score datasets, viz. NIST BSSR1, BIOCOP2008, and MIT LL Trimodal, and investigate the factors which might influence the effectiveness of imputation. Our studies reveal three key observations: (1) Imputation is preferable over not imputing missing scores, even when the fusion rule does not require complete score data. (2) Balancing the classes in the training data is crucial to mitigate negative biases in the imputation technique towards the under-represented class, even if it involves dropping a substantial number of score vectors. (3) Multivariate imputation approaches seem to be beneficial when scores between modalities are correlated, while univariate approaches seem to benefit scenarios where scores between modalities are less correlated.
Paper Structure (10 sections, 3 figures, 7 tables, 1 algorithm)

This paper contains 10 sections, 3 figures, 7 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Background
  3. Approach
  4. Datasets
  5. Missing Score Simulations
  6. Imputation
  7. Results
  8. Missing Simulation versus Naturally Occurring Missing
  9. Conclusion and Discussion
  10. Future Work

Figures (3)

  • Figure 1: Diagram of iris and ocular images cropped from ocular region in the BIOCOP2008 dataset images (RGB facial images top, NIR ocular images bottom).
  • Figure 2: Estimated TMR at FMR=0.1% for 50% incomplete score vectors found in the unbalanced version of the training data.
  • Figure 3: Estimated TMR at FMR=0.1% for 50% incomplete score vectors trained on the balanced training set.