The distributed recognition of patterns in time series data using the Patterns software is presented. Patterns is a general-purpose data analysis tool that finds patterns in data and represents the patterns as objects. The patterns can be processed by different applications (including expert systems and database techniques) to make decisions.
The Patterns software comes with an expert system for writing rules about the patterns using the information embodied in the objects. The expert system offers standard features, including a backward chaining rule interpreter and a report generator. The rule language is unique, however, as it integrates pattern recognition operations with signal processing and symbolic reasoning. The Patterns software also offers a programmable API for integrating pattern recognition capabilities with the data analysis tools of other third-party vendors.
At the center of the Patterns architecture is a distributed application receiving data from heterogeneous sources and sending the data to applications for analysis, display, closed-loop control, or data archiving.
A user interface (a template editor) is provided in order to describe the pattern of interest as a grouping of template primitives. These templates, included in a library, can be used as building blocks in other templates. The description is hierarchical (a parse tree). Patterns is based on the syntactic pattern recognition approach. However, start and end times and statistical attributes for each pattern object are included.
This paper announces an important software product and outlines some real problems that it solves. The main audience is engineers. They will find an interesting description of the subject, but no details.