Shape Similarity Index for Time Series based on Features of Euclidean Distances Histograms

Shape similarity between time series is quantified based on descriptive statistical features from histograms of Euclidean distances. The histogram shows the statistical distribution of point to point differences between two time series, using this representation of the compared time series, a shape similarity index was designed. The index quantifies the differences between histograms with respect to an ideal similarity histogram that corresponds to a reference time series, this index allows the ordering of a set of time series based on their similarity with respect to this time series. A benchmark set of time series with known similarities and dissimilarities was used in order to evaluate the performance of the proposed index. The experimental results show that this index is able to order and differentiate with good resolution the time series, by their similarity with respect to the reference time series