Anticipatory Fall Detection in Humans with Hybrid Directed Graph Neural Networks and Long Short-Term Memory

2025-09-01

AI Summary

This study addresses early fall detection by modeling the dynamic evolution from stable gait → transitional state → fall. We propose a decoupled dual-task framework: (1) a Dynamic Graph Neural Network (DGNN) classifies real-time skeletal sequences into three gait states—stable, transitional, or falling; and (2) an LSTM module predicts future joint trajectories to enhance temporal awareness. Crucially, our approach is the first to systematically characterize and identify the transient pre-fall transitional state, moving beyond conventional binary (fall/non-fall) classification. Evaluated on OUMVLP-Pose and URFD datasets, our model outperforms standalone DGNN and state-of-the-art methods, achieving new SOTA performance in both gait-state classification accuracy and motion trajectory prediction error. The framework delivers interpretable, robust monitoring of the critical transitional phase—enabling timely, actionable interventions for intelligent fall-prevention assistance systems.

Technology Category

• Intelligent Robots: State Estimation
• Planning, Routing, and Scheduling: Activity and Plan Recognition
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Detecting and preventing falls in humans is a critical component of assistive robotic systems. While significant progress has been made in detecting falls, the prediction of falls before they happen, and analysis of the transient state between stability and an impending fall remain unexplored. In this paper, we propose a anticipatory fall detection method that utilizes a hybrid model combining Dynamic Graph Neural Networks (DGNN) with Long Short-Term Memory (LSTM) networks that decoupled the motion prediction and gait classification tasks to anticipate falls with high accuracy. Our approach employs real-time skeletal features extracted from video sequences as input for the proposed model. The DGNN acts as a classifier, distinguishing between three gait states: stable, transient, and fall. The LSTM-based network then predicts human movement in subsequent time steps, enabling early detection of falls. The proposed model was trained and validated using the OUMVLP-Pose and URFD datasets, demonstrating superior performance in terms of prediction error and recognition accuracy compared to models relying solely on DGNN and models from literature. The results indicate that decoupling prediction and classification improves performance compared to addressing the unified problem using only the DGNN. Furthermore, our method allows for the monitoring of the transient state, offering valuable insights that could enhance the functionality of advanced assistance systems.

Problem

• Predicting human falls before occurrence using hybrid models
• Analyzing transient state between stability and impending falls
• Decoupling motion prediction and gait classification for accuracy

Innovation

• Hybrid DGNN and LSTM for fall prediction
• Decoupled motion prediction and gait classification
• Real-time skeletal features from video input