MHEALTH (Mobile Health) Wearable Sensor Dataset

Overview

The MHEALTH (Mobile Health) dataset was designed to benchmark human behavior analysis techniques using multimodal wearable sensing for mobile health applications.[web:124][web:130][web:153]

Data Collection

• Data were collected from 10 volunteers with diverse profiles while they performed 12 annotated physical activities, such as walking, jogging, running, cycling, standing, sitting, and jumping.[web:124][web:130][web:153]
• Three synchronized sensor units were placed on the chest, right wrist, and left ankle of each subject.[web:124][web:130][web:153]
• All sensing modalities were recorded at 50 Hz, which is sufficient to capture typical human movements.[web:124][web:130][web:153]

Signals and Variables

• Each unit records tri-axial acceleration, angular velocity (gyroscope), and magnetic field orientation, providing rich kinematic information for three body parts.[web:124][web:130][web:153]
• The chest unit additionally provides 2‑lead ECG, enabling basic heart monitoring and analysis of exercise effects on cardiac activity.[web:124][web:130][web:153]
• Data for each subject are stored in a dedicated log file (e.g., mHealth_subject<ID>.log) with rows corresponding to time samples and columns to sensor channels plus activity labels.[web:124][web:130][web:148]

Use Cases

• Wearable-based human activity recognition and behavior analysis with multiple body locations.[web:124][web:127][web:151]
• Mobile health applications combining motion and ECG for exercise monitoring and basic arrhythmia screening research.[web:124][web:148][web:150]
• Evaluation of multimodal sensor fusion methods for robust HAR on low-power IoT wearables.[web:124][web:149][web:151]

License and Terms

The UCI Machine Learning Repository notes that its datasets, including MHEALTH, are made available under a Creative Commons Attribution 4.0 International (CC BY 4.0) license, allowing sharing and adaptation with proper credit.[web:70][web:136][web:154]