Latency-information theory and applications: Part II. On real-world knowledge aided radar
Erlan H. Feria · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2008
In this second of a multi-paper series latency-information theory (LIT), the integration of information theory with its time dual, i.e., latency theory, is successfully applied to DARPA's knowledge aided sensor signal processing expert reasoning (KASSPER) program. LIT encapsulates the concept of the time dual of a lossy source coder, i.e., a lossy processor coder. A lossy processor coder is a replacement for a signal-processor. This lossy processor coder is faster, simpler to implement, and yields a better performance than the original signal-processor when the processor input appears in a highly compressed-decompressed lossy fashion. In particular, a lossy clutter covariance processor (CCP) architecture is investigated that has successfully replaced KASSPER's originally advanced lossless CCP and enabled its SAR imagery prior knowledge to be highly compressed-decompressed. This result is illustrated with a typical SAR image which is compresseddecompressed by a factor 8,172. Using this image and under severely taxing environmental disturbances outstanding detections are achieved with the lossy CCP. Furthermore, this result is derived with a lossy CCP that is at least five orders of magnitude faster and significantly simpler to implement than the corresponding lossless CCP whose SINR detection performance is nevertheless unsatisfactory. As a final comment it is also observed that LIT illuminates biological system studies since it provides a lossy mechanism that explains how outstanding detections may be arrived at by biological systems that use highly lossy compressed prior knowledge, e.g., when a human expertly detects a face seen only once before even though that face cannot be accurately described prior to such new viewing.