Activity recognition dataset - HCI gestures dataset

Daniel Roggen, Wearable Computing Laboratory, ETH Zurich
droggen@gmail.com
Initial documentation: 10.03.2009
Last updated: 06.09.2010

Description

This dataset contains 5 classes, recorded with an 8 acceleration sensors USB sensors placed on the right lower and upper arm.

It was recorded by Kilian Förster to investigate the effect of sensor displacement on activity recognition performance [Förster09].

In addition motion-jacket was recorded (not included in this dataset - available in original dataset - contact us for more informations).

Availability

This dataset can be freely used in publications provided the following paper is cited: [Förster09].

Sensors

8 3D acceleration sensors are placed on the arm used to do the gestures, at regular intervals.
Sample rate: 96Hz.

Activities

Gestures were performed vertically in 'freehand way' and also against a styrofoam support to limit gesture variability. The gestures are:

Triangle, pointing up
Square
Circle
Infinity
Triangle, pointing down

All gestures were done clockwise (for infinity the first right loop done clockwise).

1 Subject
10+ repetitions of freehand gesture
60+ repetitions of guided gesture

Files

Dataset is available in two versions: guided and freehand.

The dataset in dataset_usb_hci_dtc.mat contains two variables: dataset_usb_hci_freehand_dtc for freehand movement in DTC format; dataset_usb_hci_guided_dtc for guided movement in DTC format.

This format is: acceleration = dataset_xxx{sensor}{activity}{instance}

with sensor=0...23 are the sensor axis (8 3-axis sensors = 8x3 = 24 axis). Sensor node number is mod(s,3). For the exact position please contact us.

The original processed and corrected dataset is in usb_hci_freehand.mat and usb_hci_guided.mat. In this format, the sensor are in the order: 3, 4, 18, 19, 20, 27, 29 31

Sensor placement

References

[Förster09] K. Förster, D. Roggen, G. Tröster. Unsupervised Classifier Self-Calibration through Repeated Context Occurences: Is there Robustness against Sensor Displacement to Gain? In Proc. 13th {IEEE} Int. Symposium on Wearable Computers (ISWC 2009), pages 77-84, 2009