NetherlandsForensicInstitute/NFI_FARED_IMU

This is the README file for the dataset Netherlands Forensic Institute: Forensic Activity Recognition Dataset (NFI_FARED), published as a part of the paper "Hi-OSCAR: Hierarchical Open-set Classifier for Human Activity Recognition.". Two forms of data were collected: Digital Traces from iPhones worn on the subjects' bodies, and raw sensor signals from body-worn Inertial Measurement Units (IMUs). This dataset and README refers to the IMU data. The Digital Trace data is available here.

NFI_FARED contains accelerometer, gyroscope, and air pressure data from 14 participants (8 male, 6 female), aged 26.6 ± 8.8 years old. 2 IMU devices were worn, one on the dominant wrist and the other on the back. For further details on the data collection procedure, please refer to the paper.

All subjects (pp) are numbered 01-15 (subject 14 is not existent), with data stored in their respective folder /pp<SUBJECT NUMBER>/Sensor gegevens/verwerkte sensor gegevens NFI/ . Inside the folder the .csv files contain the actual sensor data. The naming convention of the files is: <DATE>_exp<EXPERIMENT NUMBER>_<final|freeliving>_pp<SUBJECT NUMBER>.csv . Files containing <final> are from the controlled experimental conditions, <freeliving> are from the unmonitored periods.

The .csv files contain 17 columns.

• Column 0: index
• Column 1: timestamp
• Columns 2-4: accelerometer readings of back-mounted IMU in each axis (x, y, and z) in $g$
• Columns 5-7: gyroscope readings of back-mounted IMU (x, y, z) in degrees/s
• Column 8: air pressure of back-mounted IMU
• Columns 9-11: accelerometer readings of arm-mounted IMU (x, y, and z) in $g$
• Columns 12-14: gyroscope readings of arm-mounted IMU (x, y, z) in degrees/s
• Column 15: air pressure of arm-mounted IMU
• Column 16: Activity Label

Activity labels are as follows: standing, sitting, walking, running, train, car, tram, bus, cycling, stair_up, stair_down, escalator_down, escalator_up, elevator_down, elevator_up, dragging, throwing, punching, kicking . Rows labeled no activity are from untracked moments in the recording sessions e.g. walking from the office where the IMUs were attached to the starting location. There are no assurances on the activities performed (or not performed) in these periods.

Subjects 3, 4, and 5 were not able to complete the full set of experiments, so their data is supplemented by subjects 11, 12, and 15 who completed the remaining sessions. Python scripts for processing the .csv files can be found on the project GitHub .

Citation

If you wish to use this dataset in your research please cite:

@article{10.1145/3770681,
author = {McCarthy, Conor and Quirijnen, Loes and van Zandwijk, Jan Peter and Geradts, Zeno and Worring, Marcel},
title = {Hi-OSCAR: Hierarchical Open-set Classifier for Human Activity Recognition},
year = {2025},
issue_date = {December 2025},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
volume = {9},
number = {4},
url = {https://doi.org/10.1145/3770681},
doi = {10.1145/3770681},
journal = {Proc. ACM Interact. Mob. Wearable Ubiquitous Technol.},
month = dec,
articleno = {199},
numpages = {27},
keywords = {Human Activity Recognition (HAR), Out-of-Distribution Detection, Wearable Sensors}
}

Contact

For questions regarding the data processing, Hi-OSCAR paper, and/or project GitHub please contact Conor McCarthy: c.t.mccarthy@uva.nl

For questions regarding the data collection and forensic applications please contact Jan Peter van Zandwijk: j.p.van.zandwijk@nfi.nl