In this paper, we propose dimensionality reduction representation of multi-scale time series histograms, which is performed based on the multi-scale histograms. It is a faster and efficient way to pre-select time sequence in a database and leads to reduce the need of time sequence comparisons when answering similarity queries. A new metric distance function MD ( ) that consistently lower-bounds WED and also satisfies the triangular inequality is also presented and based on it, we construct the Slim-tree index structure as the metric access method to answer similarity queries. We also extend it to subsequence matching and presented a MSST index structure.
展开▼
