Accelerated queries of time series databases

摘要

It has been previously proposed that a query to a database of time signals can be accelerated by searching over partial data in Fourier space. It is proposed here that such queries can be accelerated further by employing composite Fourier filtering. In particular, many fractional power filters are trained on sets of vectors from the database. In single comparisons information about the entire set of vectors is available. Query times are shortened mainly due to the ability to examine several vectors in the database simultaneously.