Feature Extraction Of Underwater Target Based On Auditory Features

Target recognition is the key technology to underwater operation. And the technology offeature extraction is crucial technology of it, so it is significant to research underwater targetfeature extraction technology.Vector hydrophone can not only obtain pressure and particle velocity information butalso anti isotropic noise. Sonar technician can recognize underwater targets by listening andobserving, but it is easy to be affected by external environment, psychological factors andothers. It will lead to an inaccurate judgment, so it is significant for auditory feature extractiontechnology combined with vector hydrophone in underwater targets recognition.The main research contents of this thesis are as follows:1. Radiation noise characteristics of ship is analyzed. The scalar and vector signal forunderwater targets are modeled and simulated.2. The feature based on wavelet packet energy of underwater target signals is extracted.The related theories of wavelet packet and the feature extraction are introduced and thefeature extraction method of wavelet packet energy is researched through simulation andexperiment.3. Feature based on Mel Frequency Cepstral Coefficients (MFCC) of underwater targetsignals is extracted. Acoustic target feature extraction method of MFCC is raised. The conceptof Mel Differential Cep-strum Coefficient is introduced. The feature extraction method basedon MFCC is researched through simulation and experiment.4. Feature based on Moore loudness model and principal component analysis ofunderwater target signals is extracted. The related theories of Moore loudness model andprincipal component analysis are stated. The feature extraction method is researched throughsimulation and experiment.5. BP neural network model and its general principals for design are presented. And theclassification results of above chapters by using BP neural network classifier are researched.The optimized and modified methods of MFCC are proposed.This paper through the above study, research the feasibility of auditory feature extractionmethod of acoustic vector signals to apply to target identification technology in low SNR...