Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

First International StepUP Competition for Biometric Footstep Recognition: Methods, Results and Remaining Challenges

Robyn Larracy, Eve MacDonald, Angkoon Phinyomark, Saeid Rezaei, Mahdi Laghaei, Ali Hajighasem, Aaron Tabor, Erik Scheme

TL;DR

This paper introduces the First International StepUP Competition for Biometric Footstep Recognition and the UNB StepUP-P150 dataset to benchmark verification under real-world variability. Competitors deployed end-to-end deep learning approaches, with top methods leveraging GRM-based hyperparameter optimization, TI-MFBO, and ECCO on a R(2+1)D backbone, achieving a best EER of 10.77% on a withheld test set. While performance is strong under familiar footwear and speeds, generalization to unseen footwear remains a major challenge, as demonstrated by substantial errors when personal shoes differ from training data. The work provides a practical benchmark and highlights footwear variability as a critical direction for future research in deploying robust footstep biometrics.

Abstract

Biometric footstep recognition, based on a person's unique pressure patterns under their feet during walking, is an emerging field with growing applications in security and safety. However, progress in this area has been limited by the lack of large, diverse datasets necessary to address critical challenges such as generalization to new users and robustness to shifts in factors like footwear or walking speed. The recent release of the UNB StepUP-P150 dataset, the largest and most comprehensive collection of high-resolution footstep pressure recordings to date, opens new opportunities for addressing these challenges through deep learning. To mark this milestone, the First International StepUP Competition for Biometric Footstep Recognition was launched. Competitors were tasked with developing robust recognition models using the StepUP-P150 dataset that were then evaluated on a separate, dedicated test set designed to assess verification performance under challenging variations, given limited and relatively homogeneous reference data. The competition attracted global participation, with 23 registered teams from academia and industry. The top-performing team, Saeid_UCC, achieved the best equal error rate (EER) of 10.77% using a generative reward machine (GRM) optimization strategy. Overall, the competition showcased strong solutions, but persistent challenges in generalizing to unfamiliar footwear highlight a critical area for future work.

First International StepUP Competition for Biometric Footstep Recognition: Methods, Results and Remaining Challenges

TL;DR

This paper introduces the First International StepUP Competition for Biometric Footstep Recognition and the UNB StepUP-P150 dataset to benchmark verification under real-world variability. Competitors deployed end-to-end deep learning approaches, with top methods leveraging GRM-based hyperparameter optimization, TI-MFBO, and ECCO on a R(2+1)D backbone, achieving a best EER of 10.77% on a withheld test set. While performance is strong under familiar footwear and speeds, generalization to unseen footwear remains a major challenge, as demonstrated by substantial errors when personal shoes differ from training data. The work provides a practical benchmark and highlights footwear variability as a critical direction for future research in deploying robust footstep biometrics.

Abstract

Biometric footstep recognition, based on a person's unique pressure patterns under their feet during walking, is an emerging field with growing applications in security and safety. However, progress in this area has been limited by the lack of large, diverse datasets necessary to address critical challenges such as generalization to new users and robustness to shifts in factors like footwear or walking speed. The recent release of the UNB StepUP-P150 dataset, the largest and most comprehensive collection of high-resolution footstep pressure recordings to date, opens new opportunities for addressing these challenges through deep learning. To mark this milestone, the First International StepUP Competition for Biometric Footstep Recognition was launched. Competitors were tasked with developing robust recognition models using the StepUP-P150 dataset that were then evaluated on a separate, dedicated test set designed to assess verification performance under challenging variations, given limited and relatively homogeneous reference data. The competition attracted global participation, with 23 registered teams from academia and industry. The top-performing team, Saeid_UCC, achieved the best equal error rate (EER) of 10.77% using a generative reward machine (GRM) optimization strategy. Overall, the competition showcased strong solutions, but persistent challenges in generalizing to unfamiliar footwear highlight a critical area for future work.
Paper Structure (12 sections, 5 figures, 3 tables)

This paper contains 12 sections, 5 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Datasets
  3. Training Samples
  4. Reference and Probe Samples
  5. Competition
  6. Competition Task
  7. Management and Starter Kit
  8. Evaluation Metrics
  9. Participation
  10. Submitted Solutions
  11. Results and Discussion
  12. Conclusion

Figures (5)

  • Figure 1: Overview of the collection procedure and example footstep recordings: (a) a participant walking across the pressure sensor-instrumented platform in a pair of their own personal shoes, (b) a series of footstep pressures captured during a pass across the platform, and (c) an example of an extracted and pre-processed footstep recording, shown at several time points throughout the stance.
  • Figure 2: Detection error tradeoff (DET) curve for the top five solutions. Circle markers represent model performance at the competitor's submitted threshold, while EERs are denoted by each curve's intersection with the dotted line (FNMR = FMR).
  • Figure 3: Performance (EER) for the top five submitted models across different walking speed and shoe type conditions. The Preferred Speed, Personal Shoe #1 probe samples (top left corner) include conditions 'seen' in the reference set, while the others are 'unseen' conditions.
  • Figure 4: Overlap in misclassified probe samples for the top three solutions. There were 650 especially challenging footsteps that were misclassified by all three teams.
  • Figure 5: Examples of probe footsteps that were misclassified by all three of the top solutions, shown alongside reference samples for the claimed identities. Panels (a) and (b) depict false matches, and (c) and (d) depict false non-matches. Note: the 3D (time, height, width) footstep samples are shown here as 2D peak pressure images for visualization.