Detecting Motifs Under Uniform Scaling


Dragomir Yankov, Eamonn Keogh, Jose Medina, Bill Chiu, Victor Zordan
Dept. of Computer Science & Eng.
University of California, Riverside,
USA
{dyankov, eamonn, medinaj, bill, vzb}@cs.ucr.edu

 

Supporting page for submission: Kr322_Yank.pdf(KDD07)

 

Data sets used in the experimental evaluation:

        Motifs in Brain Activity Time Series

 

Dataset(s):

Dataset C, Dataset D, Dataset E (smoothed versions, used in this paper)

 

The original time series can be obtained from the web site of the Epileptology Clinic in Bonn University

 

 

Description:

A thorough description of the data and the collection process can be found in:

 

 

 

Andrzejak RG, Lehnertz K, Rieke C, Mormann F, David P, Elger CE (2001) "Indications of nonlinear deterministic and finite dimensional structures in time series of brain electrical activity: Dependence on recording region and brain state", Phys. Rev. E, 64, 061907

 

 

 

The data represent brain wave recordings from epileptic patients. The time series in Dataset E are recorded during epileptic attacks, while Dataset C and Dataset D contain recordings from seizure-free periods. The above data sets are obtained from the original time series collected by Andrzejak et al., by running a low pass moving average filter of size ten data points.

        Motion-capture Motifs

 

Dataset(s):

Mocap data

 

 

Description:

This is a collection of 75 motion-capture sequences with duration of 10sec-30sec.


The motions captured are martial arts movements - kicks, blocks, punches and retracting movements. The time series that we extract from the data comprise the z-coordinate of the sensor attached to the left arm of the actor. For ease of the evaluation, in the experiments we first (up/down)sample all time series to the same length of 1200 data points.

 

 

Demo:

The video clip shows a demonstration of the actor performing a blocking movement. The motifs detected with both Euclidean distance and with uniform scaling are depict.

 

 

 

 

        Projectile Shapes

 

Dataset(s):

 Arrowheads

 

 

Description:

The data set contains 1231 time series of length 251. The original images for the examples used throughout the paper are also included.

The archive also contains background information on evolutionary archeology, and how some details in the arrowheads shape are being used by researchers in the field. A centroid shape extraction approach is visually demonstrated.  

 

 

Examples: