A Lightweight and Explainable Hybrid Deep Learning Model for Wearable Sensor-Based Human Activity Recognition

• Published in: IEEE sensors journal Vol. 25; no. 12; pp. 22618 - 22628
• Main Authors: Tokas, Pratibha, Semwal, Vijay Bhaskar, Jain, Sweta
• Format: Journal Article
• Language: English
• Published: New York IEEE 15.06.2025; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accuracy; Adaptation models; Band-pass filters; Datasets; Deep learning; Explainable artificial intelligence; Feature extraction; Human activity recognition; human activity recognition (HAR); Human motion; inertial measurement unit (IMU) sensor; Machine learning; Muscles; Pathology; Real time; Rehabilitation; Robustness; Sensors; Signal processing; Smartphones; surface electromyography (sEMG) signal; Telemedicine; Training; wearable sensor; Wearable sensors; Wearable technology

Human activity recognition (HAR) is critical for rehabilitation and clinical monitoring, but robust recognition using wearable sensors (e.g., sEMG or IMU) remains challenging due to signal noise and variability. We propose X-LiteHAR, a lightweight, explainable hybrid deep learning framework for real...