Signet-Software Tools for Signal Identification Using Neural Networks

We are developing a software signal processing workbench named SIGNET that simplifies exploratory analysis of multi-channel time series data. We have demonstrated, for the first time, the feasibility of building a signal- processing system around an object-oriented database OODB. This provides a graphical means for users to create, compare, and manipulate complex data structures while maintaining system wide understanding of these structures. This understanding enables the system to provide database queries by content, data subset extraction with retention of important relationships. traceable self- documenting data. insurance that only appropriate data is fed to signal processing functions, etc. The end result is that users have a high degree of flexibility to manipulate data while data integrity and validity is protected. Over the course of the project, we completed a detailed system design, evaluated existing database technologies and chose an OODB upon which to build SIGNET. We built a prototype that implemented the essential framework of SIGNET and provided a platform with which to test the basic technical issues underlying our design. Signal review and exploratory signal analysis software was enhanced for incorporation into the SIGNET framework. We have also tested and analyzed the prototype and have found that the major drawback to our initial design was the speed of system response. lie major factors causing this have been identified and speed-up solutions have been designed. We conclude that OODB technology provides a powerful and appropriate framework to model the data and processes that are used in exploratory multidimensional signal-processing applications. We are determining the commercial viability of developing the prototype into a full commercial system. Signal, Processing, Software, Neural, Networks