Accelerometer Data for Energy Harvesting During Walking Estimation

  • Christopher Beach (Contributor)
  • Alex Casson (Contributor)



Accelerometer data supporting Inertial Kinetic Energy Harvesters for Wearables: The Benefits of Harvesting at the Foot, this dataset contains accelerometer data from participants walking on a treadmill at a variety of speeds with sensors on the wrist, hip, ankle and foot.
If using this data, please cite: C. Beach, A. J. Casson, "Inertial Kinetic Energy Harvesters for Wearables: The Benefits of Harvesting at the Foot," IEEE Access, 2020 (doi:
Analysis code for this paper is available at

This repository includes both the raw data collected by the Axivity AX3 sensors (the raw CWA files and the resampled data in CSV format) and versions of the data that has been trimmed into multiple records corresponding to each speed of the treadmill (in pickled format). Note there is no participant P1.

Accessing the raw data:
The folders P2 – P13 contain the untrimmed data from the sensors.
Sample rate: 100 Hz, units: g
Participants were instructed to walk on a treadmill (LifeSpan TR1200i) as close as possible to how they would normally walk, while the speed of the treadmill was controlled by the experimenter. The treadmill was started at 2.4 km/h and the speed increased every 60 s by 0.1 km/h until the treadmill reached 4.3 km/h. Prior to recording each sensor went under a synchronisation procedure where all the sensors were flipped on their z-axis, causing a transition from -1g to +1g. The times of this synchronisation and the time for starting the treadmill is detailed in metadata.xlsx
Files ending in .cwa are in Continuous Wave Accelerometer format (a binary format) which can be processed with OmGUI software from Axivity. Files ending in .csv are these files are processed cwa files in a text format, files ending in .resampled.csv have been resampled to 100 Hz using OmGUI. These resampled files account for the fact that the AX3 sensors sample at close to 100 Hz with significant sampling jitter by resampling the data to make the sampling rate exactly 100 Hz. It is recommended to work with the sampled files.
21629: Left wrist
21704: Right wrist
31447: Left hip
32610: Right hip
32784: Left ankle
32798: Right ankle
32816: Left foot
32973: Right foot

Accessing the trimmed data:
The trimmed data can be accessed by downloading the .pkl files, which are suitable to be imported directly into Python. Each of these can be imported by running the following commands in Python:

import pickle
import numpy as np
pkl_file = open('P2.pkl', 'rb')
P2 = pickle.load(pkl_file)

The files cached_indexes.h5, cached_data.h5 and cadence_list.csv are required for use with the analysis code used in the paper and available at
Date made available16 Nov 2020
PublisherMendeley Data

Research Beacons, Institutes and Platforms

  • Manchester Environmental Research Institute
  • Manchester Institute for Collaborative Research on Ageing


  • walking
  • gait
  • accelerometer
  • wearable sensor
  • energy harvesting

Cite this