The capability to identify radar signals from enemy fighters about to launch a semi-active air-to-air missile would allow jamming before launch. If the attacking radar could be recognized within the first several pulses, the aircraft under attack would have this capability. Advent Systems proposes a nonconventional approach to the identification problem by using both intrapulse and classical parameters in a unique architecture. An important part of this architecture is a high speed neural network technology that can classify hundreds of thousands of pulses per second. Using radar waveforms obtained from Air Force data bases, and from Advent's own data base, intrapulse features will be calculated for a large number of pulses from a representative number of fighter radars and from other radar types. Statistical analyses will determine the best parameters to be used by an intrapulse sorter and a software simulation of the high speed neural network classifier. Identification accuracy and potential processing speed will be studied as a function of AMOP and FMOP feature vector domain (time, frequency and wavelet domains) and size (dimensionality). Advent Systems is confident of the proposed approach to this problem because of previous intrapulse parameter and neural network studies and development projects. The proposed processing concepts and high speed implementation techniques are unique capabilities of Advent's staff.
Keywords: AMOP FMOP NEURAL NETWORK DISCRIMINANT ANALYSIS SIGNAL SORTING WAVELET TRANSFORM INTRAPULSE