@article{4804,
  author = {Pit Pichappan},
  title = {Temporal Dynamics and Latent Structures of Emotional States in Human– Robot Interaction: A Markovian and Hidden Markov Model Analysis of Longitudinal Facial Emotion Sequences},
  journal = {Journal of Multimedia Processing and Technologies},
  year = {2026},
  volume = {17},
  number = {3},
  doi = {https://doi.org/10.6025/jmpt/2026/17/3/95-123},
  url = {https://www.dline.info/jmpt/fulltext/v17n3/jmptv17n3_1.pdf},
  abstract = {Understanding the temporal evolution of human emotions is critical for developing empathetic and adaptive
Human Robot Interaction (HRI) systems. While existing research predominantly focuses on instantaneous
emotion classification, it often overlooks the longitudinal dynamics and underlying latent affective structures.
This study introduces the Temporal Emotion Dynamics and Latent Regime Framework (TEDLRF) to model
the evolution of emotions as a stochastic temporal process. Utilizing a pre trained Vision Transformer for
facial emotion recognition, we analyzed longitudinal sequences of five basic emotions through Markov
Chains, Hidden Markov Models (HMM), and Dynamic Bayesian Networks.
Our analysis reveals that emotional states exhibit remarkable temporal inertia, with an Emotional Stability
Index of 0.991 and self transition probabilities exceeding 97%. Furthermore, HMM state discovery identifies
three distinct latent affective regimes: isolated anger, persistent sadness, and a flexible positive neutral
subsystem. Survival analysis further demonstrates significant heterogeneity in emotional persistence, where
sadness forms highly stable, long duration attractors, whereas surprise remains highly transient. Emotional
evolution is characterized by quasi stationary phases punctuated by discrete regime shifts rather than
continuous drift.
These findings indicate that observable facial expressions are manifestations of deeper, unobservable
psychological states. Consequently, we propose that next generation adaptive robots must transition from
reactive, mimetic responses to proactive, temporally aware architectures. By inferring latent emotional
regimes and anticipating affective transitions via change point detection, robots can deliver context sensitive,
empathetic behaviors, ultimately fostering more natural and effective human AI collaboration.},
}