Integrated Vehicle Classification System Using Wavelet/Neural network Processing of Acoustic Emissio

CYTEL proposes to develop a novel system for robust, high performance identification of ground vehicles from acoustic emissions. Hallmarks of this approach include the identification of an optimal wavelet decomposition for use in extracting robust multidimensional features from the acoustic emission. These wavelet-based features will be processed by additional feature post processing algorithms and coupled to a programmable, multilayer, neural network for vehicle identification. A series of algorithm variations are to be developed during the Phase I program. These variations will be used to construct a number of vehicle identification systems. All systems will be validated on a set of customer supplied ground vehicle acoustic emission data sets. Results of all algorithm combinations will be reported. The best performing set of algorithms will serve as the core for further development in the Phase II program addressing algorithm enhancements for lower signal to noise ratios and embedding in a easily updated / extendible system for real-time operations. Benefits include development of robust, high performance vehicle identification systems based on wavelet/neural network methods which represent the next generation of signal processing approaches for developing advanced systems. In addition, the Phase I effort will identify the specific combination of algorithms which will lend themselves to embedded real-time operation, leading to deployable, passive surveillance systems for both military and commercial markets.