StepGAN: Gait Feature Extraction Using Generative Adversarial Networks for Footstep Recognition

Pressure recordings of footsteps during walking can offer a convenient biometric recognition method for applications in security, forensic analysis, and health monitoring. However, footsteps can exhibit high variability due to a complex interplay of internal and external factors, posing a challenge for recognition systems. To address this issue, this study employed generative adversarial networks with a second discriminator and triplet loss to extract features from high-resolution foot pressure images. By mapping footstep data with different footwear conditions to a shared domain using barefoot pressure, the proposed StepGAN feature extractors significantly improved balanced accuracies from 93.3-95.7% to 96.8-98.0% for verification of 20 individuals with support vector machine classification. This improvement was evident even for users and conditions not included during network training, which highlights the potential of deep-generative models to learn distinctive and generalizable footstep representations. Future studies are recommended to expand these ideas to other factors that contribute to variability. Several potential research directions have been identified.