# Beyond Appearance: Transformer-based Person Identification from Conversational Dynamics

arXiv: https://arxiv.org/abs/2510.04753 · HTML: https://mlsanity.com/paper/2510.04753/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 7 of its 25 references that are in the map.
- First posted on arXiv 6 Oct 2025.

## Builds on (most cited first, 7 of 7)

- [Sapiens: Foundation for Human Vision Models](https://mlsanity.com/paper/2408.12569/) (2024)
- [Whole-Body Human Pose Estimation in the Wild](https://mlsanity.com/paper/2007.11858/) (2020)
- [Skeleton-based Action Recognition via Spatial and Temporal Transformer Networks](https://mlsanity.com/paper/2008.07404/) (2020)
- [Automatic Speech Recognition using Advanced Deep Learning Approaches: A survey](https://mlsanity.com/paper/2403.01255/) (2024)
- [Comprehensive Survey: Biometric User Authentication Application, Evaluation, and Discussion](https://mlsanity.com/paper/2311.13416/) (2023)
- [Joint Person Identity, Gender and Age Estimation from Hand Images using Deep Multi-Task Representation Learning](https://mlsanity.com/paper/2303.15263/) (2023)
- [Person Recognition using Facial Micro-Expressions with Deep Learning](https://mlsanity.com/paper/2306.13907/) (2023)

## Built on by (most cited first, 2 of 2)

- [Micro-Expression-Aware Avatar Fingerprinting via Inter-Frame Feature Differencing](https://mlsanity.com/paper/2604.23247/) (2026)
- [FinInvest-GTCN: Explainable Graph-Temporal-Causal Modeling for Risk-Aware Investment Decision Optimization](https://mlsanity.com/paper/2606.28933/) (2026)

## Abstract (arXiv)

This paper investigates the performance of transformer-based architectures for person identification in natural, face-to-face conversation scenario. We implement and evaluate a two-stream framework that separately models spatial configurations and temporal motion patterns of 133 COCO WholeBody keypoints, extracted from a subset of the CANDOR conversational corpus. Our experiments compare pre-trained and from-scratch training, investigate the use of velocity features, and introduce a multi-scale temporal transformer for hierarchical motion modeling. Results demonstrate that domain-specific training significantly outperforms transfer learning, and that spatial configurations carry more discriminative information than temporal dynamics. The spatial transformer achieves 95.74% accuracy, while the multi-scale temporal transformer achieves 93.90%. Feature-level fusion pushes performance to 98.03%, confirming that postural and dynamic information are complementary. These findings highlight the potential of transformer architectures for person identification in natural interactions and provide insights for future multimodal and cross-cultural studies.