PAMAP2 Physical Activity Monitoring Dataset

Overview

The PAMAP2 Physical Activity Monitoring dataset includes real sensor data from 9 participants performing a mix of everyday and sports activities while instrumented with body-worn inertial measurement units (IMUs) and a heart rate monitor.[web:125][web:128][web:138]

Data Collection

• Each subject performed up to 18 activities such as walking, running, cycling, climbing stairs, ironing, vacuum cleaning, and playing soccer.[web:125][web:128]
• Three wireless IMUs were placed on the dominant wrist, chest, and dominant ankle, and an additional device recorded heart rate.[web:125][web:128][web:152]
• IMU signals were sampled at 100 Hz, while heart rate was recorded at approximately 9 Hz.[web:128][web:140][web:152]

Signals, Format, and Files

• Each IMU provides tri-axial accelerometer, gyroscope, and magnetometer readings, plus temperature, giving 52 raw sensory attributes in total when combined with heart rate.[web:128][web:152]
• Data are stored in space-separated .dat files, one file per subject per session (protocol or optional), where each line has a timestamp, activity label, and the 52 sensor values.[web:128][web:138][web:142]
• Missing sensor values are indicated with NaN, and the readme file documents the activity codes and column order in detail.[web:128][web:142]

Use Cases

• Research on human activity recognition and activity intensity estimation using multi-sensor wearables.[web:125][web:128][web:155]
• Evaluation of segmentation, feature extraction, and deep learning pipelines on high-frequency wearable IoT data.[web:125][web:128][web:152]
• Benchmarking sensor fusion and domain adaptation methods across subjects and activity types.[web:128][web:140][web:152]

License and Terms

The official readme states that the dataset is freely available for academic research with no legal constraints on using the data for scientific purposes, provided that users cite the recommended references.[web:142][web:152]