Projektgruppe Visual Analytics - Medieninformatik und Multimedia ...
Projektgruppe Visual Analytics - Medieninformatik und Multimedia ...
Projektgruppe Visual Analytics - Medieninformatik und Multimedia ...
Sie wollen auch ein ePaper? Erhöhen Sie die Reichweite Ihrer Titel.
YUMPU macht aus Druck-PDFs automatisch weboptimierte ePaper, die Google liebt.
Derivative Dynamic Time Warping<br />
Eamonn J. Keogh † and Michael J. Pazzani ‡<br />
58 Kapitel 3 Multitouch<br />
1 Introduction<br />
hauptsächlich darum, zweidimensionale Sequenzen miteinander zu vergleichen. Dies<br />
wird hauptsächlich bei der Spracherkennung verwendet, lässt sich aber auch für die<br />
Erkennung Time von series Gesten are a auf ubiquitous Multitouch-Eingabegeräten form of data occurring verwenden. in virtually every scientific<br />
discipline. A common task with time series data is comparing one sequence with another.<br />
Um eine In some Geste domains mit einer a very anderen simple distance vergleichen measure, zu such können, as Euclidean müssendistance die Bewegungen will suffice. jeder<br />
Geste However, zunächst it is auf often einthe zweidimensionales case that two sequences Muster have the heruntergerechnet approximately the <strong>und</strong> same gespeichert overall<br />
werden. component Trenntshapes, man die but Daten these shapes der horizontalen do not line up <strong>und</strong> in X-axis. vertikalen Figure Bewegung 1 shows this auf, with soakann<br />
mansimple zwei Diagramme example. In erstellen. order to find Hierthe befindet similarity sichbetween dann auf such dersequences, X-Achse die or as Zeit a <strong>und</strong><br />
auf der preprocessing Y-Achsestep jeweils before die averaging horizontale them, oder we must vertikale "warp" Bewegung. the time axis of Imone ersten (or both) Prozess<br />
werden sequences die X-Achse to achieve dera gespeicherten better alignment. Geste Dynamic <strong>und</strong> die time X-Achse warping (DTW), der soeben is a technique ausgeführten<br />
for efficiently achieving this warping. In addition to data mining (Keogh & Pazzani 2000,<br />
Bewegung angeglichen (s. Abb. 3.16). Dies erfolgt über eine Matrix, welche die<br />
Yi et. al. 1998, Berndt & Clifford 1994), DTW has been used in gesture recognition<br />
Distanzen der Y-Werte enthält. Durch Linearisierung (Abb. 3.17) der Distanzen werden<br />
(Gavrila & Davis 1995), robotics (Schmill et. al 1999), speech processing (Rabiner &<br />
die Werte angeglichen.<br />
Juang 1993), manufacturing (Gollmer & Posten 1995) and medicine (Caiani et. al 1998).<br />
A)<br />
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70<br />
Abbildung Figure 1: An 3.16: example Angleichung of the utility of der dynamic X-Achse time warping. beim Dynamic A) Two sequences Time-Warping that represent [Keo01] the Y-<br />
axis position of an individual’s hand while signing the word "pen" in Sign Language. The sequences were<br />
recorded on two separate days. Note that while the sequences have an overall similar shape, they are not<br />
aligned in the time axis. A distance measure that assumes the i th point on one sequence is aligned with i th<br />
point on the other will produce a pessimistic dissimilarity. B) DTW can efficiently find an alignment<br />
between the two sequences that allows a more sophisticated distance measure to be calculated.<br />
B)<br />
4<br />
† Department of Information and Computer Science University of California, Irvine,<br />
California 92697 USA eamonn@ics.uci.edu<br />
‡ pazzani@ics.uci.edu<br />
0 5 10 15 20 25 30<br />
1<br />
m<br />
0 5 10 15 20 25 30<br />
j<br />
1 w1<br />
1<br />
w3<br />
w2<br />
…<br />
Figure 3: An example warping path.<br />
Abbildung 3.17: Die Abbildung zeigt die Abstände<br />
2.1 Constraining zwischen den the classic beidendynamic Sequenzen, time warping welchealgorithm<br />
anschließend<br />
The problem<br />
linearisiert<br />
of singularities<br />
werden [Keo01]<br />
was noted at least as early as 1978 (Sakoe, & Chiba<br />
1978)). Various methods have been proposed to alleviate the problem. We briefly review<br />
them here.<br />
1) Windowing: (Berndt & Clifford 1994) Allowable elements of the matrix can be<br />
restricted to those that fall into a warping window, |i-(n/(m/j))| < R, where R is a<br />
positive integer window width. This effectively means that the corners of the<br />
matrix are pruned from consideration, as shown by the dashed lines in Figure 3.<br />
Others have experimented with various other shaped warping windows (Rabiner et<br />
al 1978, Tappert & Das 1978, Myers et. al. 1980). This approach constrains the<br />
maximum size of a singularity, but does not prevent them from occurring.<br />
2) Slope Weighting: (Kruskall & Liberman 1983,Sakoe, & Chiba 1978) If equation<br />
i<br />
wK<br />
n