Inaccuracies of Shape Averaging Method Using Dynamic Time Warping for Time Series Data

Shape averaging or signal averaging of time series data is one of the prevalent subroutines in data mining tasks, where Dynamic Time Warping distance measure (DTW) is known to work exceptionally well with these time series data, and has long been demonstrated in various data mining tasks involving shape similarity among various domains. Therefore, DTW has been used to find the averageshape of two time series according to the optimal mapping between them. Several methods have been proposed, some of which require the number of time series being averaged to be a power of two. In this work, we will demonstrate that these proposed methods cannot produce the realaverage of the time series. We conclude with a suggestion of a method to potentially find the shape-based time series average.

[1]  Aaron E. Rosenberg,et al.  Speaker independent recognition of isolated words using clustering techniques , 1979, ICASSP.

[2]  Eamonn J. Keogh,et al.  Iterative Deepening Dynamic Time Warping for Time Series , 2002, SDM.

[3]  Eamonn J. Keogh,et al.  Everything you know about Dynamic Time Warping is Wrong , 2004 .

[4]  E. Caiani,et al.  Warped-average template technique to track on a cycle-by-cycle basis the cardiac filling phases on left ventricular volume , 1998, Computers in Cardiology 1998. Vol. 25 (Cat. No.98CH36292).

[5]  A. Corradini,et al.  Dynamic time warping for off-line recognition of a small gesture vocabulary , 2001, Proceedings IEEE ICCV Workshop on Recognition, Analysis, and Tracking of Faces and Gestures in Real-Time Systems.

[6]  D. H. Lange,et al.  Modeling and estimation of single evoked brain potential components , 1997, IEEE Transactions on Biomedical Engineering.

[7]  O. Meste,et al.  Integral shape averaging and structural average estimation: a comparative study , 2005, IEEE Transactions on Signal Processing.

[8]  L. Gupta,et al.  Nonlinear alignment and averaging for estimating the evoked potential , 1996, IEEE Transactions on Biomedical Engineering.

[9]  Eamonn J. Keogh,et al.  An Enhanced Representation of Time Series Which Allows Fast and Accurate Classification, Clustering and Relevance Feedback , 1998, KDD.

[10]  B. Ray,et al.  An Interweaved HMM/DTW Approach to Robust Time Series Clustering , 2006, 18th International Conference on Pattern Recognition (ICPR'06).

[11]  Paul S. Bradley,et al.  Refining Initial Points for K-Means Clustering , 1998, ICML.

[12]  R. Coifman,et al.  Local feature extraction and its applications using a library of bases , 1994 .

[13]  Edward A. Fox.  Digital Libraries: Implementing Strategies and Sharing Experiences, 8th International Conference on Asian Digital Libraries, ICADL 2005, Bangkok, Thailand, December 12-15, 2005, Proceedings , 2005, ICADL.

[14]  Lawrence R. Rabiner,et al.  A modified K-means clustering algorithm for use in isolated work recognition , 1985, IEEE Trans. Acoust. Speech Signal Process..

[15]  V. Mor-Avi,et al.  Signal averaging helps reliable noninvasive monitoring of left ventricular dimensions based on acoustic quantification , 1994, Computers in Cardiology 1994.

[16]  Philip Chan,et al.  Toward accurate dynamic time warping in linear time and space , 2007, Intell. Data Anal..

[17]  Stan Salvador,et al.  FastDTW: Toward Accurate Dynamic Time Warping in Linear Time and Space , 2004 .

[18]  Waleed H. Abdulla,et al.  Cross-words reference template for DTW-based speech recognition systems , 2003, TENCON 2003. Conference on Convergent Technologies for Asia-Pacific Region.

[19]  Gareth J. Janacek,et al.  Clustering Time Series with Clipped Data , 2005, Machine Learning.

[20]  Eamonn J. Keogh,et al.  Multimedia Retrieval Using Time Series Representation and Relevance Feedback , 2005, ICADL.

[21]  Aaron E. Rosenberg,et al.  Speaker-independent recognition of isolated words using clustering techniques , 1979 .

[22]  Paul R. Cohen,et al.  Using Dynamic Time Warping to Bootstrap HMM-Based Clustering of Time Series , 2001, Sequence Learning.