Research On Technology Of Passive Identification Of Underwater Target

The recognition of passive underwater acoustic target is one hot and difficult research inmost countries in the world,specially in the military, as ship sound hidden ability graduallyincreased, this technology will become increasingly urgent demand. National naval authoritieshave attached great importance to research and innovation in the field of passive underwateracoustic target recognition. Based on the analysis on the basis of the ship radiated noisecharacteristics, an overall architecture for passive underwater acoustic target recognition hasbeen proposed in this paper. This architecture is consists of data preparation, feature extraction,classification and recognition and information fusion of four parts.Data preparation section is based on the current understanding of the ship radiated noisecharacteristics, use the tools such as filters to analog produce the simulation signal whoserelative features is consistent with the real ship radiated noise in a laboratory environment.Simulation ship radiated noise signal has all the characteristics and information needed by therecognition system, get easy, and the amount of data can be freely controlled, to some extent, thishas solve the problem of difficulties of obtaining data in underwater acoustic experiments. Thispaper analyzes the causes and characteristics of each component in the ship radiated noise, basedon it, a mathematical model of the ship radiated noise simulation is established, and a method ofusing specific frequency response filter to simulate continuous ship radiated noise spectrum isproposed. Simulation results indicate that the specific frequency response method has generatedthe noise spectrum which is more realistic than the spectrum generated by other methods, it haslaid a good foundation for the verifying of the performance of the recognition system.The classifier is defined as which can determine a sample belong to which categoryaccording to the input mode, realize the category of framing. This paper recommends the LVQneural network classifier which is based on artificial neural network. As a development of aself-organizing competition network, this classifier not only inherited the competitivecharacteristics of the network, but also increased the supervised learning algorithm, in order toachieve the classification and identification of mentors. This paper introduces the classifier to thearea of underwater acoustic target classification and identification, LVQ network is established,the simulated ship radiated noise data is used to extract feature vectors, the network is trainedand tested. The results show that the network model can quickly identify the target, and thehigh-dimensional feature vector can be identified more accurate than traditional BP neuralnetwork with higher recognition rate.Information fusion can make a reasonable integration of the information from differentplatforms, to achieve better recognition effect. According to the experimental results, BP neuralnetwork and LVQ network has a good ability to identify the line spectra features and waveletenergy features. Thus, this paper introduces both of the two pieces of identification to therecognition system, so that each identification can play a role, fusion identifiable information in the decision-making level, experiments show that the decision-making level information fusioneffectively improve the recognition rate of the whole system.