Energy Entropy Detection And Cesptrum Analysis Of The Underwater Signal

The ocean development has become a heat issue in the current world, and the marine research attracts more and more attention recently. Whether for the military applications or the economic development, the target detection and recognition are all needed to pay more attention. And the application of the vibration and noise reduction on the ship reduced the radiated noise of the underwater targets, it become more important to search for much effective target detection and identification methods. So it has great theoretical significance and application values. The main work of this paper is using the feature of radiation noise of the target to reach the purpose of the target detection and identification. The paper can be divided into four aspects as follows:Firstly, according to the mechanism of the radiated noise of the underwater targets, the simulation of the noise is worked out and getting the simulation data of the vector information of the radiated noise of the target.Secondly, using the detection algorithm based on the entropy of the energy of the noise radiated by the underwater target. The energy entropy detection algorithm based on wavelet transform and empirical mode is comprehensive presented and simulated.And using the simulation signal calculating the detection performance of the two types of detection algorithms, and compared with the classical energy detection algorithm.Thirdly, we use the cepstrum to extract the feature of the radiated noise of the underwater target. Making the feature extraction of two kinds of simulation data using the linear prediction cepstrum(LPCC), Mel frequency cepstrum(MFCC) and MFCC based on linear prediction(LP-based MFCC), the result show that these feature can be as the feature vector for target identification.Finally, the support vector machine for target classification has been studied mainly. We get the feature vector of the scalar and vector properties two kinds of measured data using LPCC, MFCC and LP-based MFCC and get the recognition result of the scalar and vector properties.