I want to understand Timeseries shape similarity algorithm ( Shape-based distance aka SBD). I can't understand the statistics behind it and why it is better than DTW or other similarity measure. I'm trying to read this paper http://www1.cs.columbia.edu/~jopa/Papers/PaparrizosSIGMOD2015.pdf but can't understand anything in this regard. Any explanation with some example is really appreciated.
I started with learning about cross-correlation, normalized cross-correlation, Auto-Correlation, Normalization and Time Shift.
I couldn't understand the concept of sliding cross-correlation which is used to identify shape based similarity. What is "sliding" in this type of correlation and what difference it's making?