Application of the chaos, fractal and wavelet theories to the feature extraction of passive acoustic signal

Firstly, the chaos performance and the nature of nonstationary of the aerodynamic noise arediscussed. Then, the distributions of the maximum Lyapunov exponent λ1, the fractal dimension andthe energy distribution indifferent wavelet scale space of the noise of the tank and helicopter are testedand analyzed statistically. Finally, these feature vectors and BP neural network are used to classifythe realistic data of tank and helicopter. Results show that the classification system can achieve astatisfactory classification accuracy in different SNR.