The original data include 10 subjects, each perform 10 gestures 10 times. The gesture acquisition device is a Nintendo Wiimote remote controller with built-in three-axis accelerometer. Each subject performs a set of gestures multiple times. Classes are based on gestures (see class labels below). Note that data are shuffled and randomly sampled so that instances across datasets are not sychronized by dimension or subject. Time series are of different lengths. There are no missing values.

1. poteg: pick-up
2. shake: shake
3. desno: one move to the right
4. levo: one move to the left
5. gor: one move to up
6. dol: one move to down
7. kroglevo: one left circle
8. krogdesn: one right circle
9. suneknot: one move toward the screen
10. sunekven: one move away from the screen

This data is acceleration in x-axis dimension.