Probing Identity-Specific Motion Signatures: A Controlled Diagnostic Study

2026-07-03

AI Summary

This study investigates whether modern video models effectively leverage identity-specific motion cues for person recognition rather than relying solely on static appearance. To this end, the authors introduce BALLER120, a rigorously controlled benchmark dataset comprising free-throw sequences from professional basketball players, with inputs provided as RGB frames, silhouettes, and skeletal representations. Through attention visualization and ablation studies, they demonstrate that identity-specific motion micro-patterns indeed exist and are learnable by models. Moreover, when appearance information is deliberately suppressed, the models maintain competitive accuracy despite reduced input information, exhibit significantly improved robustness to variations in visual appearance, and become more attuned to subtle, individualized motion characteristics.

Abstract

Identity recognition (e.g., person, animal re-identification) has traditionally relied heavily on static appearance cues. Yet motion--consistent, individual-specific dynamics--can provide a complementary and potentially more robust signature, especially when appearance is weak or variable. This raises a fundamental question: when identity-specific motion cues are clearly present, to what extent do modern video models use them for recognition? To investigate this question, we conduct a systematic diagnostic study and introduce BALLER120, a controlled benchmark of 120 professional basketball players performing free-throws. By focusing on the same multi-phase action across individuals, BALLER120 reduces action-level variation and identity-correlated acquisition biases, enabling fine-grained analysis of identity-specific kinematic patterns. We find that modern video models can predict identity accurately from RGB videos, but often rely on static appearance cues such as faces and jersey regions, even when informative motion cues are available. Strikingly, when appearance is suppressed through silhouette-only or skeleton-only inputs, the same model architectures shift toward motion micro-patterns (e.g., foot placement and elbow bending). Despite containing less visual information, appearance-suppressed representations achieve competitive accuracy and stronger robustness to appearance shifts. Our qualitative analyses further show that appearance-suppressed models attend to distinctive motion patterns across individuals. Overall, our study demonstrates that identity-specific motion signatures are present, informative, and learnable, but modern video models may overlook them in favor of easier static shortcuts unless appearance cues are explicitly suppressed.

Problem

• identity recognition
• motion signatures
• video models
• appearance cues
• kinematic patterns

Innovation

• identity-specific motion
• controlled benchmark
• motion signatures
• appearance suppression
• video model diagnostics