Using signature files for querying time-series data

Abstract

This paper describes our work on a new automatic indexing technique for large one-dimensional (1D) or time-series data. The principal idea of the proposed time-series data indexing method is to encode the shape of time-series into an alphabet of characters and then to treat them as text. As far as we know this is a novel approach to 1D data indexing. In this paper we report our results in using the proposed indexing method for retrieval of real-life time-series data by its content.