Acoustic sensors are the technology of choice for Unattended Ground Sensors and Smart Munitions, like WAM. A key function of these sensors is to classify and identify target vehicles. Existing algorithms need to be improved to permit operation in low signal to noise conditions, against convoys and other multi-target scenarios and over a wider range of operating conditions. In addition, these systems must be field reprogrammable. Current classifiers extract spectral features and process them in a pattern matching algorithm against stored feature data. Their performance is limited when these features change with vehicle perating conditions. No single collection of reference features can cover the wide range of possible operating conditions. Many types of operating conditions and vehicle categories can be recognized using the time-frequency characteristics of their signatures. This SBIR proposes to develop a system to extract time-frequency features and analyze them using a knowledge-based (or expert) system that will determine what set of features and reference data to use in the pattern matching process. This adaptive matching process will greatly expand the conditions over which targets can be classified. Algorithms will be tested against existing target signature data.Acoustic sensors with vehicle identification capabilities can have applications in a) vehicle traffic monitoring, b) airport noise monitoring systems, c) security Systems. Adaptive signature recognition algorithms can be used in a) machinery condition monitoring, b) quality control and c) medical applications.
Keywords: Acoustic Sensors Target Classification Unattended Ground Target Identification Statistical Pattern
