Phase space engineering for neurodynamic target identification

The authors propose and describe a new concept in automated radar target identification from a single look (coherent broadband echo) based on neural net models. They view a neural net as a multidimensional nonlinear dynamic system capable of exhibiting powerful collective computational and signal processing functions that are fully and therefore best described by their phase-space behavior in terms of terminal, periodic, or strange attractors and associated basins of attraction. They suggest that a central issue in understanding and applying neural nets today is finding ways for imparting to a net distinct phase-space behavior that gives rise to desired functions. This phase-space engineering approach represents a totally new paradigm in signal processing that originates from known universal features of biological information processing in the nervous system. The authors present initial results illustrating one possible method for applying the phase-space engineering concept to the long-standing problem of recognizing airborne targets from a single look.<>