Welcome to the page of Mobility Dataset, developed by the team of
Md. Atiqur Rahman Ahad, Senior Member, IEEE.
Professor
Dept. of Electrical and Electronic Engineering (EEE)
[former Dept. of Applied Physics, Electronics & Communication Engineering]
University of Dhaka, Bangladesh.
Specially Appointed Associate Professor, Dept. of Media Intelligent
Osaka University, Japan.
About the Dataset
The University of Dhaka Mobility Dataset (DU-MD / MD) is a sensor-based human action/activity recognition (HAR) dataset.
It has 10 activity classes, and 5000 observations from 50 subjects recorded using wrist-mounted sensors
embracing accelerometry. The dataset is continually updated. It exhibits sufficient statistical diversity in physiological parameters
and a noteworthy correlation between similar activities with coveted quantitative and qualitative features,
suitable for training machine learning models. On the other hand, the wrist-mounted approach parallels the
future commercial scenarios. The sensors are specially made in the University of Tokyo.
This work is partially based on the results of a project commissioned by the New Energy and Industrial Technology Development Organization (NEDO).
Keywords:
Mobility dataset, Human Actiion Recognition (HAR), Activities of Daily Living (ADL), Wearable sensors, accelerometer
Action Classes:
- Walking,
- Sitting,
- Lying,
- Jogging,
- Staircase climbing,
- STaircase down,
- Standing,
- Falling via unconsciousness,
- Falling via heart attack, and
- Falling via slipping while walking.
Principle:
Accelerometry using a wrist-mounted data logger; The newly developed open-source Internet of
Things (IoT) kit from IIS, University of Tokyo, known as the Trillion Node Engine Project, was used to assemble
the data logger [Field Assemblable and Con gurable Open-Source IoT Device Platform,
Univ. Tokyo, Tokyo, Japan, 2018.].
Features:
- Three types of realistic falls are taken into account for fall
detection.
- Signals are sampled at 30 Hz to ensure detailed ADL waveform.
- Kruskal-Wallis (K-W) one-way ANOVA was performed on 10 walking ADL from a random test subject. Each ADL contains 101 sample points.
At the 0.05 level, the populations are not significantly different, which justifies that similar ADL are statistically
coherent and uniform.
- Kolmogorov-Smirnov (K-S) test was performed on age, weight and height of 25 subjects to ensure sufficient statistical diversity.
- No bias was observed in weight and height, whilst a small bias was observed in age, which was corrected by augmenting the significance level of
the test to 0.1.
Download
Download the segmented (Window Length: 101) data: Segmented Raw Data
Subjects' information who volunteered: Volunteers' Info
Related Publications of this dataset (please cite these if you explore or refer)
- S.S. Saha, S Rahman, MJ Rasna, TB Zahid, AKMM Islam, and Md Atiqur Rahman Ahad, "Feature Extraction, Performance Analysis and System Design using the DU Mobility Dataset", IEEE Access, IEEE, Vol. 6, pp. 44776-44786, 2018. [Impact Factor: 3.557]
DOI: 10.1109/ACCESS.2018.2865093 Open Access / Download
- S.S. Saha, Shafizur Rahman, Miftahul Jannat Rasna, A.K.M. Mahfuzul Islam, and Md Atiqur Rahman Ahad, "DU-MD: An Open-Source Human Action Dataset for Ubiquitous Wearable Sensors", Joint 7th Int'l Conf. on Informatics, Electronics & Vision, & 2nd Int'l Conf. on Imaging, Vision & Pattern Recognition, Japan, 2018.
[Achieved Best Paper Award Candidate].
- S.S. Saha, S. Rahman, M. J. Rasna, T. Hossain, S. Inoue, and Md Atiqur Rahman Ahad, "Supervised and Neural Classifiers for Locomotion Analysis",
Adjunct Proceedings of the 2018 ACM International Joint Conference on Pervasive and Ubiquitous Computing and the 2018 International Symposium on Wearable Computers Proceedings, ACM, Singapore, 2018.
DOI: 10.1145/3267305.3267524
Contact
- Contact Email: [for Saha] swapnilsayansaha [_ATmark_] ieee DOT org.
[for Ahad] atiqahad [_ATmark_] du DOT ac DOT bd, or atiqahad [_ATmark_] yahoo DOT com
- Postal Address:
Md. Atiqur Rahman Ahad, Ph.D., SMIEEE
Professor
Dept. of Electrical and Electronic Engineering (EEE)
University of Dhaka
Dhaka - 1000
Bangladesh
©2018, Swapnil Sayan Saha, Md. Shafizur Rahman, Miftahul Jannat Rasna, A.K.M. Mahfuzul Islam, Md Atiqur Rahman Ahad
This work is partially based on the results of a project commissioned by the New Energy and Industrial Technology Development Organization (NEDO).
The work is in part supported by FAB Lab DU and IIS, The University of Tokyo.
|
