Explainable Fall Detection for Elderly Care via Temporally Stable SHAP in Skeleton-Based Human Activity Recognition

2026-04-16 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Medical Devices & Health Technology · Depth: Expert, quick

Summary

A new framework for explainable fall detection in elderly care, combining an efficient LSTM model with T-SHAP, has been proposed to address the temporal instability of existing post-hoc explainability methods. This approach stabilizes SHAP-based feature attributions over time by applying a linear smoothing operator, preserving theoretical guarantees while reducing high-frequency variance. Evaluated on the NTU RGB+D Dataset, the framework achieved 94.3% classification accuracy with an inference latency under 25 milliseconds, meeting real-time requirements for mid-range hardware. Quantitative analysis using perturbation-based faithfulness metrics showed T-SHAP improved explanation reliability over standard SHAP (AUP: 0.91 vs. 0.89) and Grad-CAM (0.82), consistently highlighting biomechanically relevant motion patterns like lower-limb instability and spinal alignment changes.

Key takeaway

For AI Engineers developing explainable AI systems for healthcare, this research demonstrates that incorporating temporal stability into post-hoc explanation methods like SHAP is critical. You should consider T-SHAP or similar temporal aggregation strategies to ensure that your model's explanations are consistent and trustworthy for clinicians, especially in time-sensitive applications like fall detection where reliable insights into biomechanical patterns are paramount.

Key insights

Temporally stable SHAP explanations enhance reliability for skeleton-based fall detection in elderly care.

Principles

• Explanation stability is crucial for clinical trust.
• Linear smoothing improves temporal attribution consistency.

Method

The method combines an LSTM model with T-SHAP, a temporally aware aggregation strategy that applies a linear smoothing operator to SHAP attribution sequences to stabilize feature attributions over contiguous time windows.

In practice

• Deploy real-time fall detection on mid-range hardware.
• Use T-SHAP for reliable clinical decision aids.

Topics

• Explainable AI
• Fall Detection
• Skeleton-Based Human Activity Recognition
• Temporally Stable SHAP
• LSTM

Best for: AI Engineer, Research Scientist, AI Scientist, Machine Learning Engineer, Computer Vision Engineer

Related on AIssential

• Explainable Fall Detection for Elderly Care via Temporally Stable SHAP in Skeleton-Based Human Activity Recognition — Temporally stable SHAP attributions improve reliability for explainable fall detection in elderly care.
• Explainable AI in Production: A Neuro-Symbolic Model for Real-Time Fraud Detection — Integrating explanations directly into model architecture enables real-time, deterministic, and low-latency explainability.
• Quantifying Explanation Consistency: The C-Score Metric for CAM-Based Explainability in Medical Image Classification — The C-Score quantifies explanation consistency in CAM methods, revealing model instability beyond classification metrics.
• Explainable AI based ensemble model for the identification of Schizophrenia prodromal phase — An XAI-integrated ensemble ML model accurately predicts Schizophrenia prodromal symptoms for early clinical intervention.
• Towards Rigorous Explainability by Feature Attribution — Existing SHAP scores for ML explainability are flawed and misleading; rigorous alternatives are necessary.
• When Confidence Lacks Concepts: Interpretable OOD Detection via Representation Perturbations — Interpretable OOD detection can be achieved by analyzing prediction stability under concept-conditioned semantic perturbations.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.