Dataset for Mobile Phone Sensing Based Fall Detection

This is a labelled dataset of mobile phone integrated accelerometer and gyroscope sensor data, in which falls and activities of daily life (ADL) were simulated as realistically as possible by martial artists.

5 different mobile phone devices were used for the recordings:
Samsung Galaxy S4
Samsumg Galaxy S5 Mini
Sony Z3
Google Nexus 4
Google Nexus 5
Sampling Frequency: 5Hz, for both accelerometer and gyroscope sensors

Simulation of falls and ADL
We recorded eight simulations, seven incorporating a fall and one simulation with ADL. The simulations were performed by five martial artists in parallel. Each simulation was repeated five times. The simulations were repeated with a second group of martial artists.
Each martial artist was equipped with a belt bag in which the mobile device was put into.

Dataset Content
13 Tables
DEVICE_ALL_ACCELERATION ... acceleration data of all devices
DEVICE_ALL_ORIENTATION ... orientation data of all devices
DEVICE_{modelname}_ACCELERATION ... acceleration data of a single device
DEVICE_{modelname}_ORIENTATION ... orientation data of a single device
MANUAL_LABELS ... manually labelled activities

Datatypes
Acceleration is stored as a three dimensional vector indicating acceleration along each device axis, not including gravity. All values have units of m/s^2. Addtionally, the total acceleration from the vector is stored. Formula: squared root of (x^2 + y^2 + z^2)

Orientation is stored as three dimensional vector of angles azimuth, pitch and roll. All values are angles in degrees.

Azimuth, angle between the magnetic north direction and the y-axis, around the z-axis (0 to 359). 0=North, 90=East, 180=South, 270=West
Pitch, rotation around x-axis (-180 to 180), with positive values when the z-axis moves toward the y-axis.
Roll, rotation around y-axis (-90 to 90), with positive values when the x-axis moves toward the z-axis.
Note: This definition is different from yaw, pitch and roll used in aviation where the X axis is along the long side of the plane (tail to nose).

Timestamps are in milliseconds (unix epoch time).

Stats
159300 acceleration records
159300 orientation records
492 labels
~ 6 six hours of fall simulations
~ 4 hours of ADL simulations

ACKNOWLEDGEMENTS

The Knowledge Technologies Institute is part of the Faculty of Computer Science and Biomedical Engineering at the Graz University of Technology.

The Know-Center is funded within the Austrian COMET Program under the auspices of the Austrian Ministry of Transport, Innovation and Technology, the Austrian Ministry of Economics and Labor and by the State of Styria. COMET is managed by the Austrian Research Promotion Agency FFG.

The DALIA project (AAL-2012-5-249) is co-financed under the Ambient Assisted Living Joint Programme of the European Commission (www.aal-europe.eu) and the National Funding Agencies of Austria, Belgium, Netherlands and Switzerland.

The authors also wish to thank the martial artists of the Ninpo Bujutsu Club Graz, for simulating so many falls.