MoME: Estimating Psychological Traits from Gait with Multi-Stage Mixture of Movement Experts

2025-10-06

AI Summary

This study addresses the emerging and challenging task of predicting psychological traits from gait sequences. To tackle difficulties in modeling 2D pose sequences and the weak supervision inherent in psychological labels, we propose a Multi-stage Mixture of Motion Experts (MoME) architecture: it partitions the walking cycle into four motion-complexity stages, each served by a lightweight expert network, with dynamic fusion guided by a task-aware gating mechanism; notably, it introduces the first hierarchical mixture-of-experts design that jointly models auxiliary tasks—including identity, gender, and BMI—to strengthen psychological representation learning. Trained end-to-end on the PsyMo benchmark, MoME achieves a subject-level weighted F1-score of 44.6%, substantially outperforming prior methods. This result provides the first systematic empirical validation of gait-based psychological inference, demonstrating both its feasibility and effectiveness.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior

Application Category

• User Modeling, Personalization and Recommendation: Psychology-informed user models and recommender systems
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI

Abstract

Gait encodes rich biometric and behavioural information, yet leveraging the manner of walking to infer psychological traits remains a challenging and underexplored problem. We introduce a hierarchical Multi-Stage Mixture of Movement Experts (MoME) architecture for multi-task prediction of psychological attributes from gait sequences represented as 2D poses. MoME processes the walking cycle in four stages of movement complexity, employing lightweight expert models to extract spatio-temporal features and task-specific gating modules to adaptively weight experts across traits and stages. Evaluated on the PsyMo benchmark covering 17 psychological traits, our method outperforms state-of-the-art gait analysis models, achieving a 37.47% weighted F1 score at the run level and 44.6% at the subject level. Our experiments show that integrating auxiliary tasks such as identity recognition, gender prediction, and BMI estimation further improves psychological trait estimation. Our findings demonstrate the viability of multi-task gait-based learning for psychological trait estimation and provide a foundation for future research on movement-informed psychological inference.

Problem

• Estimating psychological traits from human gait sequences using pose data
• Developing hierarchical architecture to model multi-stage movement complexity
• Improving multi-task prediction by integrating auxiliary biometric recognition tasks

Innovation

• Hierarchical multi-stage mixture of experts architecture
• Lightweight expert models extract spatio-temporal gait features
• Task-specific gating modules adaptively weight experts