SBIR-STTR Award

The goal of this proposal is to improve the figure-of-merit of the focus of attention stage of automatic target recognition (ATR) by developing a novel hierarchical multi-resolution prescreening architecture addressing: (A) new sensory input filtering methods that highlight target discriminating features and discard nonrelevant features; and (B) a more complete characterization of targets, decoys and clutter by modeling both the representations at each scale and across scales by means of learned generative nonlinear models. the combination of improvements (A) and (B) is represented in our Wavelet Pattern Detection (WPD-ATR) approach incorporating the prescreening of input to a nonlinear adaptive system (neural net). The WPR-ATR is designed fr ultrahigh resolution sensors/radars - e.g., TeraHertz sensors/radars - and testable either in conjunction with such sensors/radars as a hardware-in-the-loop demonstration of as simulation/emulation on BMDO's VDHTB. Anticipated Benefits and

Potential Commercial Applications:
Accurate ATR is of fundamental importance to all arms of the military. The battlefield is a fast-paced environment which can only be serviced by accurate ATR. Equally, the biomedical field is inundated with records of images which presently have to be scanned by highpaid personnel. Development of automatic scanners is thus of importance.

This Phase II proposal addresses the hardware implementation of a wavelet filter set unique to BSEI and permitting fast automatic target recognition (ATR) by pattern detection (WAVERS). WAVERS is based on the IIR filter principles and is thus faster (requiring less operations) than current FIR wavelet filter sets. WAVERS is hierarchical and requires no decimation of the sensory input to achieve a multiresoultion analysis. It has wide commercial application in (a) wireless communications, achieving sharper filter roll-off with less than half the multiplication workload of FIR filters, and also reliability of transmission in the presence of jammers and noise over set time-frequency bandwidths, as the filters are orthogonal; and in (b) biomedicine, achieving fast automatic image and time series processing and information compression. The Phase II work addresses a hardware implementation of the Phase I work - either in FPGA, DSP or ASIC form, up to the manufacturing stage. Specifically, the implementation is guided by I/O requirements of the BMDO sensor contractor, Raytheon Missile System, Tucson AZ, who provided the I/O specifications for the filter set. The WAVERS filter set will significantly improve both the resolution and speed of ATR for BMDO and other DoD services. WAVERS is scalable and modular and will also find wide application in the wireless communications and biomedical processing markets