查看文章

We propose a novel indexing structure, L-index, for searching similar patterns in large time sequences efficiently. L-index is a sequence of line segments which compactly approximates the original data sequence. A line segment is used to approximate a subsequence of the original data sequence until the accumulated error exceeds a threshold value. By scanning the L-index, we can quickly find approximate answers of similarity queries. An important feature of the L-index is that approximate answers are guaranteed to include exact answers. it does not produce false dismissal. Our experimental result shows that the L-index does not require much storage area and process similarity queries efficiently.