New comprehensive review tackles the challenge of making AI-powered activity recognition systems more transparent and trustworthy for real-world healthcare and smart home applications.

Human activity recognition (HAR) using deep learning has improved performance on sensor data but created 'black box' models that lack transparency, hindering real-world deployment

Explainable AI (XAI) is emerging as critical for making HAR systems more transparent, reliable, and human-centered across healthcare monitoring, assistive living, and smart environments

The review provides a unified framework separating conceptual dimensions of explainability from algorithmic mechanisms across wearable, ambient, physiological, and multimodal sensing platforms

A new mechanism-centric taxonomy of XAI-HAR methods is presented to reduce ambiguities and clarify explanation approaches in prior research