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

2026-04-14 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Computer Vision & Pattern Recognition, Internet of Things (IoT) & Connected Devices · Depth: Expert, quick

Summary

A new lightweight and explainable framework has been developed for skeleton-based fall detection in elderly care, combining an efficient LSTM model with T-SHAP. T-SHAP is a temporally aware post-hoc aggregation strategy designed to stabilize SHAP-based feature attributions over contiguous time windows, addressing the temporal instability of existing frame-by-frame explainability methods. This framework achieves 94.3% classification accuracy on the NTU RGB+D Dataset with an end-to-end inference latency under 25 milliseconds, meeting real-time requirements for clinical monitoring. Quantitative evaluation using perturbation-based faithfulness metrics demonstrates that T-SHAP improves explanation reliability compared to 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 research scientists developing AI for clinical monitoring, this framework demonstrates that integrating temporal stability into explainability methods is crucial. You should consider T-SHAP to generate more reliable and clinically actionable insights from sequential data, especially where trust and interpretability are paramount for adoption. This approach can enhance the utility of AI in sensitive applications like elderly care.

Key insights

Temporally stable SHAP attributions improve reliability for explainable fall detection in elderly care.

Principles

• Temporal stability enhances clinical trust in AI explanations.
• Smoothing attribution sequences reduces high-frequency variance.

Method

T-SHAP applies a linear smoothing operator to SHAP attribution sequences, stabilizing feature attributions over time windows while preserving Shapley value guarantees.

In practice

• Integrate T-SHAP with LSTM for real-time fall detection.
• Use skeleton data to identify lower-limb instability.
• Prioritize explanation reliability in clinical AI systems.

Topics

• Explainable AI
• Fall Detection
• T-SHAP
• Skeleton-Based Recognition
• LSTM Models

Best for: 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 explanations enhance reliability for skeleton-based fall detection in elderly care.
• 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.
• Applied Explainability for Large Language Models: A Comparative Study — Gradient-based explainability provides stable LLM insights, while attention is efficient but less aligned with predictions.
• 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.
• 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.
• LLM Doesn't Know What It Doesn't Know: Detecting Epistemic Blind Spots via Cross-Model Attribution Divergence on Clinical Tabular Data — LLMs struggle with epistemic self-awareness on structured clinical data, but cross-model attribution divergence can detect blind spots and improve reliability.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.