josefheidler/har_ws_adults_2025-harth

NTNU Walking Speed — Thigh + Back Accelerometry, Healthy Adults

Two Axivity AX3 accelerometers (right thigh and lower back, 50 Hz) worn during a controlled overground walking protocol at three speed tiers and jogging by 24 healthy adults in Norway. Harmonized from the NTNU harth-ml-experiments release into per-subject parquet — one row per raw sample. 24 participants, 1.76 M samples, 100% labelled, 9.8 h labeled.

Source

• Paper: Logacjov, A., Bach, K., Kongsvold, A., & Mork, P. J. (2025). The performance of a machine learning model in predicting accelerometer-derived walking speed. Heliyon. https://doi.org/10.1016/j.heliyon.2025.e41625
• Raw data: https://github.com/ntnu-ai-lab/harth-ml-experiments/tree/main/adult_walking_speed
• License: CC-BY-4.0

Protocol

• Participants: 24 healthy adults (14 female, 10 male), mean age 36.1 ± 11.9 years (range 23–62), recruited at a university hospital in Trondheim, Norway.
• Sensor: Axivity AX3 triaxial accelerometer (±8 g, 50 Hz), placed on the right thigh (≈10 cm above the kneecap) and lower back (≈3rd lumbar vertebra), USB connector pointing down.
• Protocol: Overground walking in walkways at a university hospital. Each participant completed five consecutive ≈5-minute bouts at three instructed walking speed tiers (slow, moderate, fast) and one jogging bout, plus one ≈5-minute bout with gradual speed increase (slow → fast → slow). Speed was measured in real time with a CatEye Padrone bicycle speedometer mounted on a measuring wheel.

Ground-truth provenance

A trained observer followed each participant and simultaneously recorded the speedometer reading with a GoPro Hero 3+ camera. Video was annotated frame-by-frame using ANVIL software. Speed tiers were assigned post-hoc based on measured speed thresholds (slow ≤4 km/h, moderate 4.1–5.4 km/h, fast ≥5.5 km/h, jogging ≥6.5 km/h). All samples are labeled — there are no unlabeled transition windows in the release.

Schema

All timestamps are tz-aware Europe/Oslo.

column	dtype	notes
subject	string	participant alias (e.g. alert-albatross)
timestamp	datetime64[ns, Europe/Oslo]	real wall-clock time of the recording, tz-aware Europe/Oslo
thigh_acc_x	float64	right-thigh acceleration, corrected frame (see Axis orientation), units = g
thigh_acc_y	float64	right-thigh acceleration, corrected frame, units = g
thigh_acc_z	float64	right-thigh acceleration, corrected frame, units = g
back_acc_x	float64	lower-back acceleration, corrected frame (see Axis orientation), units = g
back_acc_y	float64	lower-back acceleration, corrected frame, units = g
back_acc_z	float64	lower-back acceleration, corrected frame, units = g
label	string (nullable)	activity base (walk or run)
variant	string (nullable)	walking speed tier (slow, moderate, fast); null for run
speed_kph	float64 (nullable)	approximate speed — cohort-mean from paper Table 2 (see Label vocabulary)

Example rows

subject          timestamp                       thigh_acc_x  thigh_acc_y  thigh_acc_z  back_acc_x  back_acc_y  back_acc_z  label  variant   speed_kph
alert-albatross  2020-11-02 12:12:30.649+01:00    0.969       -0.093        0.107        1.000        0.125      -0.264      walk   slow      3.1
alert-albatross  2020-11-02 12:18:43.610+01:00    1.428       -0.697        0.322        1.659       -0.135      -0.422      walk   moderate  4.9
alert-albatross  2020-11-02 12:21:11.210+01:00    0.810       -0.261       -0.117        0.734       -0.047      -0.081      walk   fast      6.1
alert-albatross  2020-11-02 12:35:10.409+01:00    2.061       -0.028       -0.040        1.254       -0.172      -0.753      run    <NA>      8.3

Label vocabulary

label and variant are stored in separate columns. speed_kph contains the cohort-mean measured speed for each tier (approximate — individual participants varied within each tier; see measured ranges below).

label	variant	meaning	speed_kph	measured range (km/h)	labeled min
run	—	jogging	8.3	7.1–9.7	114.6
walk	fast	brisk walking (≥5.5 km/h tier)	6.1	5.7–6.8	144.8
walk	moderate	moderate walking (4.1–5.4 km/h)	4.9	4.4–5.3	164.9
walk	slow	slow walking (≤4 km/h tier)	3.1	2.3–3.7	160.8

Total labeled: 585.1 min / 9.8 h across 24 participants.

Coverage

All 1,755,535 rows are labeled — this dataset has no unlabeled transition windows. Filter to labeled rows (no-op but consistent with hub convention):

df_labeled = df[df["label"].notna()]

Axis orientation

All accelerometer columns use the hub standard axis convention, shared across all datasets on this hub:

• x runs along the body segment toward the head — reads +1 g when the person stands upright, drops toward 0 when they lie down.
• y points to the person's right — positive when tilting right, negative when tilting left.
• z points forward — positive when leaning or stepping forward, negative when leaning backward.
• At rest standing upright: x ≈ +1 g, y ≈ 0, z ≈ 0.

Thigh sensor (right thigh, ≈10 cm above the kneecap, USB connector pointing down):

Native frame (standing): x=down, y=right, z=backward. Correction: negate x and z (keep y). Hub frame: x=up, y=right, z=forward.

Back sensor (lower back, ≈3rd lumbar vertebra, USB connector pointing down):

Native frame (standing): x=down, y=left, z=forward. Correction: negate x and y (keep z). Hub frame: x=up, y=right, z=forward.

The raw native-frame values are available in the source CSV files at the GitHub link above.

Harmonization notes

No participants were excluded. All 24 participants from the GitHub release are included. speed_kph values are cohort-level means from Table 2 of the paper — they are approximate. Individual participants walked at slightly different speeds within each tier (ranges documented in the Label vocabulary table above).

Use

Intended for walking-speed prediction and human activity recognition from dual-sensor thigh + back accelerometry. The dataset contains only controlled lab walking at fixed speed tiers — making it well-suited for training and evaluating walking speed classifiers but not for free-living generalization. Evaluate with leave-one-subject-out (LOSO) cross-validation across 24 participants; avoid random row-level splits which would leak temporal context across subjects.

Loading

from datasets import load_dataset

ds = load_dataset("josefheidler/har_ws_adults_2025-harth")
df = ds["train"].to_pandas()

Load a single subject:

import pandas as pd

df = pd.read_parquet(
    "hf://datasets/josefheidler/har_ws_adults_2025-harth/harmonized/alert-albatross.parquet"
)

Citation

Logacjov, A., Bach, K., Kongsvold, A., & Mork, P. J. (2025). The performance of a machine learning model in predicting accelerometer-derived walking speed. Heliyon. https://doi.org/10.1016/j.heliyon.2025.e41625