# NTNU Walking Speed: Thigh and Lower-Back Accelerometry of 24 Adults Walking at Paced Speed Bands and Jogging

## Facts

- Provider: Norwegian University of Science and Technology (Logacjov, Ludvigsen, Bach, Kongsvold, Flaaten, Lund Nilsen, Mork); harmonized by Josef Heidler
- License: MIT (https://raw.githubusercontent.com/ntnu-ai-lab/harth-ml-experiments/main/LICENSE)
- Size: 27 files
- Format: Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/josefheidler/har_ws_adults_2025-harth
- Access: public, readable without an account
- Category: Physiology
- Source: https://doi.org/10.1016/j.heliyon.2025.e42185
- Page: https://scigantic.com/archive/f4661186-2a54-5bac-82cb-50a3316f36bd
- Keywords: Physiology, Parquet, Hugging Face

## About

Accelerometry for predicting walking speed from wearable sensors, from NTNU in Trondheim. Twenty-four healthy adults (14 female, 10 male, age 23 to 62) wore two Axivity AX3 triaxial accelerometers (plus or minus 8 g, 50 Hz) on the right thigh and the lower back while walking overground in hospital walkways in five bouts of about 5 minutes: three instructed speed bands (up to 4, 4.1 to 5.4 and at least 5.5 km/h), one jogging bout above 6.4 km/h, and one bout with gradually increasing then decreasing speed. Speed was measured in real time with a bicycle speedometer on a measuring wheel, and participants were paced to the band, so labels are speed bands rather than continuous speed. The release has about 1.78 million samples and 9.9 labelled hours. Use it to train and validate gait speed estimation, a clinical vital sign, from thigh-worn accelerometers. This Hugging Face copy harmonizes the NTNU adult_walking_speed folder into one Parquet file per participant under harmonized/; load with Hugging Face datasets or pandas. NTNU publishes the data in its harth-ml-experiments repository under the MIT License. Cite Logacjov et al., Heliyon 11(2):e42185, 2025.

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("josefheidler/har_ws_adults_2025-harth")
print(ds)
```

## Data structure

File types in sampled files: Parquet

Example file: `harmonized/alert-albatross.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/f4661186-2a54-5bac-82cb-50a3316f36bd
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/f4661186-2a54-5bac-82cb-50a3316f36bd
- Over MCP: `get_schema_card` with id `f4661186-2a54-5bac-82cb-50a3316f36bd` at https://api.scigantic.com/mcp

## Related topics

- [Parquet datasets](https://scigantic.com/topics/parquet)
- [Physiology datasets](https://scigantic.com/topics/physiology)
