Application of Advanced Spectral Techniques for Ground Vehicle Acoustic Signature Classification

The goal of the proposed research is to develop robust automated detection and classification of acoustic emissions from ground vehicles. These detection algorithms will apply novel signal processing algorithms to the unique features of ground vehicles. Ground vehicles have complex, highly nonstationary acoustic signatures with rich spectral content across a broad range of frequencies. Innovative spectral techniques can be applied to these signals to extract uniquely distinctive classification clues. Prior evaluation by ORINCON of ARDEC data for tanks has identified at least three signature components that can be exploited: intercoherent harmonic line structures, broadened spectral lines, and nonstationary spectral structures. Harmonic line structures offer unique identifying information about the vehicle class. Broadened spectral lines contain subtle spectral clues that can be extracted with additional processing. Nonstationary features provide an important classification clue in multitarget environments and can be exploited through special feature-versus-time gram construction and image-based gram processing using template-based pattern matching. In addition, ORINCON has extensive experience in acoustic classification of submarines. There is a high potential for cross-discipline payoff in applying these capabilities and expertise to the ground vehicle classification problem. These advanced spectral techniques have direct applications for machinery condition monitoring, which includes piston engines for cars and trucks, turbines for jet engines, helicopter engines, and power plants, and all kinds of compressors and pumps (fuel, water, air, etc.), both reciprocating and scroll-type. This technology can also be used to monitor quasi-periodic biological processes such as heartbeat, blood flow, and heart valve action.