I wasn't sure if should post the question here, or on math.stackexchange, but here it goes. When I have to compare 2 data sequences with DTW, should I just do the algorithm on the sequences or should I first normalize the data of both of the sequences ? For example, I have two curves and both of them represent human movement. Is it okey to just do DTW and calcualte the differences between the curves ? I guess that my question is, when do you need to normalize the data before using DTW ?
1 Answer
The answer is YES, see "1.2.1 Time Series Subsequences must be Normalized" in Searching and Mining Trillions of Time Series Subsequences under Dynamic Time Warping.
http://www.cs.ucr.edu/~eamonn/SIGKDD_trillion.pdf
eamonn