Machine Learning–Powered Wearable Sensors for Gait Analysis: A Technical Validation in Healthy Subjects Toward Parkinson’s Disease Monitoring
Year: 2026
Type: article
Abstract: Parkinson's disease is characterized by progressive motor impairments that require continuous and objective monitoring beyond episodic clinical assessments. This study presents the design and technical validation of an Internet of Things-based wearable system for inertial gait monitoring, integrating a smartwatch sensing node, a mobile gateway, cloud analytics, and a web-based dashboard. We conducted a standardized experimental protocol with 40 apparently healthy adults performing three walking conditions (slow, normal, and fast), yielding 120 gait segments acquired in fixed 10-second windows. We conducted this study under controlled conditions to establish a baseline for future clinical validation in Parkinson’s disease populations. We extracted temporal, spectral, and step-related features from triaxial accelerometer signals and used them to train a supervised gait classification model based on a Random Forest algorithm, evaluated using stratified five-fold cross-validation. The proposed classifier achieved consistent discrimination performance under inter-subject variability. Although this work is limited to healthy participants, the results demonstrate the technical feasibility and reliability of the proposed system. The platform provides a technically validated foundation for future deployment in Parkinson's disease populations, supporting the development of monitoring applications and digital motor biomarkers.
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7226546944
