Time series similar pattern matching based on wavelet and dynamic time warping

The paper proposed a dynamic time warping (DTW) indexing and similar matching method of time series based on discrete wavelet transform, which reduced the dimensionality of time series by discrete wavelet transform and constructed multi-dimensional index structure by R^*-tree. The DTW lower bound and its discrete wavelet transform of query sequence were computed to form a query super-rectangle, thus the similar matching in original space based on DTW was converted to that in wavelet transform space based on Euclidian distance. It was proved that the method guaranteed no false dismissals and proposed the range query algorithm and nearest neighbor query algorithm. The result showed that it was a higher query precision and lower computing cost.