Dataset

SensHAR

wrist-worn IMU (accel/gyro) for six-activity HAR (ESP32-S3 + QMI8658A, n=30)

IMUIeee dataportCC BY 4.0Raw waveforms

Wrist-worn six-axis IMU dataset tied to the SensHAR human-in-the-loop HAR framework (Soumyajit Basak; created 2026-10-14). Custom ESP32-S3 wristband with QMI8658A IMU at 20 Hz; right-wrist accel + gyro; six labeled activities (walk, run, sit, stand, stair down/up). Study n=30 (20 generalized / 10 personalization); archive ships Combined20right.csv + S1right.csv…S10right.csv (~13.85 MB zip).

  • Published: IEEE DataPort, 2026
  • Cohort: 30 participants
  • License: CC BY 4.0
At a glance
Type
Dataset
Access
Not reported
Subjects
30
Signals
IMU
PPG sites
None
Setting
Not reported
Wrist
Yes
Raw waveforms
Yes
Tasks
HAR
Updated
Oct 16, 2026

Original description

"SensHAR is a wrist-worn inertial-sensing dataset associated with the SensHAR human-in-the-loop wearable framework for human activity recognition. The accompanying data archive contains one combined file named Combined_20_right.csv and ten participant-named files, S1_right.csv through S10_right.csv. The records include right-wrist accelerometer and gyroscope measurements, timestamp-related fields, and activity-label fields; some participant files also contain device or inference metadata. The SensHAR study describes data collection from 30 participants performing six activities2014walking, running, sitting, standing, stair descent, and stair ascent2014using a custom ESP32-S3 wristband with a QMI8658A six-axis IMU sampled at 20 Hz. In the manuscript's evaluation protocol, 20 participants provide source data for generalized model development and ten participants are held out for target-user personalization. The archive should be used only after consulting the README for its file-specific schemas and label conventions. The CSV files in the current upload do not have a uniform schema or an evident common label mapping; users should not assume that all numeric labels correspond directly to the six classes reported in the manuscript. This dataset is intended to support research on wearable inertial sensing and human activity recognition, subject to the documented limitations and verification notes."

Verbatim fromDataCite· Licence: https://creativecommons.org/licenses/by/4.0/legalcode

SensHAR: ChatPPG research summary card

Key numbers

30
Subjects
1
Reference labels
ACTIVITY
20 Hz
Released sampling rate
IMU 20 Hz (QMI8658A 6-axis)

How it works

How SensHAR works: Devices & body sites → Signals recorded → TasksDevices & body sites: ESP32-S3 wristband + QMI8658A 6-axis IMU (SensHAR) (Espressif (MCU) / QST (QMI8658A) · Wristband). PPG sites: None. Signals recorded: SensHAR, 30 subjects · 20 Hz, signals IMU. Tasks: Activity recognition.Devices & body sitesESP32-S3 wristband + QMI8658A 6-axis IMU (SensHAR): Espressif (MCU) / QST (QMI8658A) · WristbandESP32-S3 wristband +…Espressif (MCU) / QST (QMI8658A) ·…PPG sites: NonePPG sitesNoneSignals recordedSensHAR30 subjects · 20 HzxyzIMUTasksActivity recognitionActivity recognition

No model in the hub reports results on this dataset yet.

Sampling
IMU 20 Hz (QMI8658A 6-axis)
Reference labels
activity
Device roles
Study wearable (IMU)

Devices· 1

Illustrations by form factor; product photos are added as licences clear.

Models trained or evaluated on it

No model in the hub reports results on SensHAR yet. Be the first: results submitted with a paper are added here.

Cite

@misc{senshar2026,
  title = {{SensHAR: wrist-worn IMU (accel/gyro) for six-activity HAR (ESP32-S3 + QMI8658A, n=30)}},
  year = {2026},
  howpublished = {Ieee dataport},
  doi = {10.21227/31y2-7819},
  url = {https://ieee-dataport.org/documents/senshar-wrist-worn-inertial-sensor-data-human-activity-recognition}
}

History

  1. Record updated