Latent Dynamics · 2026-11-13
InertiaMPDD: Multimodal Personalized Depression Detection via Emotional Inertia-Inspired Dynamics Modeling
Changzeng Fu, Songyang Wang, Chang Gao, Xinyi Zhang, Peilin Li, Mo Meng, Jiarui Wu, Mingze Li, Hongwei Ding, Shiqi Zhao, Yong Li, Lang He
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
InertiaMPDD learns a healthy next-state predictor from non-depressed recordings using frozen pretrained multimodal representations. It compares predicted vs observed next latent states to form reference-derived descriptors: reference deviation, residual decay rate (AR(1) on residuals), and a Jacobian spectral descriptor. These proxies support standalone screening and a personalized correction to a multimodal population margin, achieving high fusion F1 on D-Vlog and F1 on LMVD, with limited standalone discrimination on AVEC2014.
Why it matters
The authors target depression screening evidence beyond label prediction under recording-level supervision by providing measurable, reference-derived behavioral dynamics proxies inspired by emotional inertia. The work separates reference informativeness from fusion benefit and evaluates context-dependent utility across multiple cohorts/datasets, using an interactive demo.
Method
- Learn a healthy next-state predictor (IIP) from non-depressed training records using pretrained multimodal representations; predict next latent state from observed history and context, then compute prediction residuals.
- Build reference-derived emotional dynamics descriptors: reference deviation (average predictive disagreement), residual decay rate (AR(1) fit to residual centered at recording median), and a Jacobian spectral descriptor (spectral radius of local Jacobian).
- Use three descriptors for standalone screening, and for personalized decision correction by adding a learned correction to a LoRA-adapted population margin (NSP) derived from multimodal latent states.
Limitation
The authors report that standalone descriptors “discriminate weakly on AVEC2014.”
Abstract (from arXiv)
Multimodal depression screening estimates risk from speech, visual behavior and language. Current methods mainly predict labels, with limited evidence of departures from a non-depressed reference. Recording-level supervision also leaves fine-grained emotional dynamics unannotated. Motivated by emotional inertia, the persistence of affect, InertiaMPDD learns healthy next-state patterns using a world-model-inspired latent predictor. Using pretrained multimodal representations, it predicts from observed history and context. Reference deviation, residual decay rate and a Jacobian spectral descriptor form reference-derived emotional dynamics descriptors, evaluated as candidate proxy biomarkers for standalone screening and personalized decision correction. Across five cohorts from four datasets, fusion reaches 98.41% F1 on D-Vlog and 84.98% on LMVD. Standalone descriptors outperform fixed additive fusion on MPDD but discriminate weakly on AVEC2014. These findings distinguish reference informativeness from fusion benefit and reveal context-dependent utility.
Related papers
- Robust EEG-based Emotion Recognition from Coarse Trial-level Labels under Intra-trial Affective Variability
- Similar Predictive Fit but Different Latent Dynamics: Characterizing Learned Dynamical Structure in Personalized Models of Brain Disorders
- Learning Multimodal Sensor Fusion with JEPA Representations for Breathing Activity and Pose Recognition
- AffectWorld: Benchmarking Emotion-Driven Future World Prediction
- Forecasting Patient Trajectories from the Electronic Health Records with Action-Conditioned World Models
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First announced by arXiv: 2026-11-13 · Indexed by Pub-AI: 2026-11-13 UTC
InertiaMPDD: Multimodal Personalized Depression Detection via Emotional Inertia-Inspired Dynamics Modeling
First announced by arXiv: 2026-11-13 · Original record ↗
- arxiv
- 2611.16647
Explicit source versions (1)
- v1 · Date not supplied
- Preprint available
Source date: 2026-11-13 · First detected: 2026-11-13 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · RSS guid=oai:arXiv.org:2611.16647v1; arxiv:announce_type=new; item pubDate (official metadata)
Abstract · arxiv
Multimodal depression screening estimates risk from speech, visual behavior and language. Current methods mainly predict labels, with limited evidence of departures from a non-depressed reference. Recording-level supervision also leaves fine-grained emotional dynamics unannotated. Motivated by emotional inertia, the persistence of affect, InertiaMPDD learns healthy next-state patterns using a world-model-inspired latent predictor. Using pretrained multimodal representations, it predicts from observed history and context. Reference deviation, residual decay rate and a Jacobian spectral descriptor form reference-derived emotional dynamics descriptors, evaluated as candidate proxy biomarkers for standalone screening and personalized decision correction. Across five cohorts from four datasets, fusion reaches 98.41% F1 on D-Vlog and 84.98% on LMVD. Standalone descriptors outperform fixed additive fusion on MPDD but discriminate weakly on AVEC2014. These findings distinguish reference informativeness from fusion benefit and reveal context-dependent utility.
