Sadiq Sani, Nirmalie Wiratunga, Kay Cooper
Robert Gordon University
Aberdeen, UK

Data Set Information:

The SELFBACK dataset contains data of 9 activity classes; 6 ambulatory activities
and 3 sedentary activities, performed by 33 participants.
Data are recorded with two tri-axial accelerometers sampling at 100Hz, mounted on
the dominant side wrist and the thigh of the participant.

**Application**
The dataset can be used for human activity recognition by developing algorithms for
pre-processing, feature extraction, sensor fusion, segmentation and classification.

** Data collection method **
Each participant performed an activity for approximately 3 minutes.
** Sensors**
Axivity AX3 3-Axis Logging Accelerometer
- sampling frequency -- 100Hz
- range -- 8g
** Activity Classes**
- Walking Upstairs
- Walking Downstairs
- Walking in slow pace
- Walking in medium pace
- Walking in fast pace
- Jogging
- Standing
- Sitting
- Lying
** Data folder **

SELFBACK dataset has three folders, two folders one for each sensor modality
named "w" for wrist and "t" for thigh and an additional folder where two sensor
modalities are merged using timestamp named "wt" for wrist and thigh.
Inside "w" and "t" folders, 9 folders can be found, one for each activity class, and
inside, there are 33 files, one file for each participant.
Inside "wt" folder, there are 297(33 X 9) files where the file name indicates the
person and the activity.

Attribute Information:

The 4 columns in the files in t and w folder is organized as follows:
1 -- timestamp
2 -- x value
3 -- y value
4 -- z value
Min value = -8
Max value = +8
The 6 columns in the files in wt folder is organized as follows:
1 -- wrist x value
2 -- wrist y value
3 -- wrist z value
4 -- thigh x value
5 -- thigh y value
6 -- thigh z value
Min value = -8
Max value = +8

Relevant Papers:

- Sani, S., Wiratunga, N., Massie, S., & Cooper, K. (2016, December).
SELFBACK--activity recognition for self-management of low back pain. In
International Conference on Innovative Techniques and Applications of Artificial
Intelligence (pp. 281-294). Springer, Cham.
- Sani, S., Massie, S., Wiratunga, N., & Cooper, K. (2017, August). Learning deep
and shallow features for human activity recognition. In International Conference
on Knowledge Science, Engineering and Management (pp. 469-482). Springer,
Cham.
- Sani, S., Wiratunga, N., Massie, S., & Cooper, K. (2017, June). kNN sampling for
personalised human activity recognition. In International conference on case-
based reasoning (pp. 330-344). Springer, Cham.
- Sani S, Wiratunga N, Massie S, Cooper K. Personalised human activity
recognition using matching networks. In International Conference on Case-
based Reasoning 2018 Jul 9 (pp. 339-353). Springer, Cham.
- Wijekoon, A., Wiratunga, N., Sani, S., Massie, S., & Cooper, K. (2018, July).
Improving kNN for Human Activity Recognition with Privileged Learning Using
Translation Models. In International Conference on Case-based Reasoning (pp.
448-463). Springer, Cham.

- Wijekoon, A., Wiratunga, N., Sani, S., & Cooper, K. (2020). A knowledge-light
approach to personalised and open-ended human activity recognition.
Knowledge-based Systems, 192, 105651.

Citation Request:

Sani, S., Wiratunga, N., Massie, S., & Cooper, K. (2016, December).
SELFBACK--activity recognition for self-management of low back pain. In?International Conference
on Innovative Techniques and Applications of Artificial Intelligence?(pp. 281-294). Springer, Cham.