This effort enhances and extends the technology foundation developed in Phase I.  The resultant technology will support (a) Parallel application of multiple classifiers to RADAR and/or EO/IR sensor data to identify probable target type/ track/location; (b) Processing of partial classification results in pattern space, for efficiency and visibility to developers to produce refined classifier results that fuse multiple sensor data streams; (c) Adaptation of classifiers and pattern-space processing to track targets given nonergodic (statistically changing) inputs; (d) Analysis of complexity and performance on several processor system architectures, with best-performance implementation on a hardware system. The Phase II research will utilize adaptive pattern selection and pattern-space processing to develop, test, and verify prototype data fusion / target classification algorithms that incorporate partially-corrupted signatures and will also produce a prototype hardware implementation.

In addition to its initial applications, the technology developed will be able to support a wide variety of pattern recognition applications for both government and industry.  These include computer-assisted diagnosis, facial recognition, speech analysis, equipment health management and camouflaged target identification.