Systematic profile/feature-based intelligence for spectral sensors

摘要

Argonne National Laboratory (ANL) has been creating a special-purpose software-engineering tool to support research and development of spectrum-output-type (chemical) sensors. The modular software system is called SAGE, the Sensor Algorithm Generation Environment and includes general-purpose signal conditioning algorithms (GP/SAGE) as well as intelligent classifiers, pattern recognizes, response accelerators, and sensitivity analyzers. GP/SAGE is an implementation of an approach for delivering a level of encapsulated intelligence to a wide range of sensors and instruments. It capitalizes on the genene classification and analysis needed to process most profile-type data. The GP/SAGE native data format is a generalized one-dimensional vector, signature, or spectrum. GP/SAGE modules form a computer-aided software engineering (CASE) workbench where users can experiment with various conditioning, filtering, and pattern recognition stages, then automatically generate final algorithm source code for data acquisition and analysis systems. SAGE was designed to free the (chemical) sensor developer from the signal processing allowing them to focus on understanding and improving the basic sensing mechanisms. The SAGE system's strength is its creative application of advanced neural computing techniques to response-vector and response-surface data, affording new insight and perspectives with regard to phenomena being studied for sensor development.