CVAP Arm/hand activity database



Data and Sensor placement:
The data in this database was recorded using a magnetic tracker (Nest of Birds) with four sensors. Each sensor gives XYZ positions and rotation matrix of each sensor. The sensors where placed in the following manner:
Sensor 0 on the chest/torso
Sensor 1 on the back side of the hand
Sensor 2 on the thumb
Sensor 3 on the index finger


Actions:
These five actions were recorded:
Push an object forward
Push an object to the side
Rotate an object 180 degrees
Move an object to the side
Grasp an object (pick up an object)

Each action is performed at 2 different heights, at 2 different positions on the table. The actions are performed from 3 different positions in front of the table and by 10 individuals.


Data:
The data base consists of 3 sets. The first set is just the recorded data. In the second set the data is labelled into 7 classes and in the third the data is labelled into 8 classes.

In the 7 classes set the actions were labelled as follows:
Push forward: Approach object – push forward – remove hand
Push aside: Approach object – push aside – remove hand
Rotate: Approach object – rotate – remove hand
Move aside: Approach object – move aside – remove hand
Grasp: approach object – grasp object – remove hand

Experiments and examination of the data suggested that a more correct labelling would be into 8 classes, since both “Rotate” and “Move aside” actions include grasping of the object. The 8 class data set was labelled as follows:
Push forward: Approach object – push forward – remove hand
Push aside: Approach object – push aside – remove hand
Rotate: Approach object – grasp object - rotate – remove hand
Move aside: Approach object – grasp object – move aside – remove hand
Grasp: approach object – grasp object – remove object


Download
The .rar file (12.1 MB) contains the data files and some Matlab files for splitting data into or 8 classes.

The content in the data files is like this:
Sensor_0:_x_y_z
39.3576 60.9898 -54.3483
matrix
0.00183105 0.142944 -0.989746
-0.958008 0.283813 0.0390625
0.286621 0.947998 0.137573
Sensor_1:_x_y_z
66.1132 59.8624 -18.2389
matrix
0.0809326 -0.301392 0.949951
0.769043 0.625 0.132446
-0.634033 0.719849 0.282227
Sensor_2:_x_y_z
62.5748 64.2826 -17.7478
matrix
-0.256226 0.0119629 0.966431
0.893188 -0.379395 0.241333
0.369263 0.925049 0.0861816
Sensor_3:_x_y_z
64.0259 63.3115 -13.1043
matrix
-0.499756 0.239136 0.832397
0.317993 0.944458 -0.0806885
-0.805664 0.224487 -0.548096
xxxxxxxxxx


This is an example of one single measurement. Each new measurement
is presented in the same format. They are separated by "xxxxxxxxxx"


Related publications are under review.


Authors

Martin Larsson, Isabel Serrano Vicente, Danica Kragic
martinla, isabel, danik  "@" nada.kth.se
Computational Vision and Active Perception lab
Centre for Autonomous System, KTH, Sweden

Ville Kyrki

kyrki "@" lut.fi
Department of Information Technology
Lappeenranta University of Technology, Finland