Description:
OBJECTIVE: Design a signal classifier that is capable of handling diverse, agile signals in an energy efficient manner in a dense signal environment. DESCRIPTION: Signals of interest (SOI) are becoming much more frequency agile, numerous, and have a low probability of intercept (LPI) by design, making signal recognition and classification much harder. The saturation of state-of-the-art Si clock speeds, increased operational tempo, and surging interest in light-weight, low-power, distributed sensors have focused attention on low latency, energy efficient digital signal processing. Re-optimization of signal classifiers is needed to allow discrimination based on agilely defined subsets of a large set of potential attributes which differentiate specific SOI. Initial ambiguity regarding which attributes are required call for multiple interpretation hypotheses to be generated, their reality probability evolved, and the branching set thinned every time a new piece of information is received. A method for backing up to still viable hypotheses when new information assigns a low probability to a previously high probability hypothesis is needed. The virtue of maintaining an evolving database of all the signals currently in the environment, as a way of eliminating the need to recategorize packeted and discretely pulsed signals, could be considered. Systems capable of differentiating receiver caused spurs and intermodulation distortion (IMD) from real signals are desirable. Improved data management efficiency is a key requirement. PHASE I: Determine the feasibility of and develop an approach, as outlined in the topic objective and description, to improve the energy efficiency of signal classification in a dense environment of agile signals. During the Phase I base, complete a proof of concept demonstration and scope a plan for Phase II. The Phase I option may be used to prove an additional functional virtue of the approach, begin work on a foreseeable hard problem, and/or refine the Phase II plan. PHASE II: Implement the Phase I approach, building out the software/firmware/hardware enablements and demonstrating their utility and energy efficiency in a simulated dense signal environment. The limiting factors on the ability of the new system to keep up in real-time with an evolving situation should be quantitatively explored. Its applicability to wideband reception of all incident signals should be evaluated. It is likely that Phase II will involve classified aspects. Demonstrate the technology developed in Phase II under more realistic settings during a possible Phase II option. PHASE III: Transition to a program of record in Phase III. PRIVATE SECTOR COMMERCIAL POTENTIAL/DUAL-USE APPLICATIONS: Under the name cognitive radios, the practice of moving signals into a currently available portion of the spectrum and back out if the owner arrives has become widely appreciated as a means of increasing capacity at minimum spectral cost. Indeed, the XiMax standard is evolving toward both center frequency and instantaneous bandwidth agility. This sort of energy efficient signal classification will also assist base stations in sorting the signal environment into paying users, non-disruptive interlopers, and spectral interlopers who are disruptive and should be pursued (e.g., thorough the civil courts). The techniques developed may also have applicability to large data set mining (e.g., in looking for purchasing trends and crowd sourcing). The more processing that can be done locally, the lower the charges will be for cloud computing and data movement.
