| Target recognition plays a significant role in modern underwater acoustic technology. Inthe process of underwater acoustic countermeasure, accurate target recognition is the base ofnot only the attack to specific target but also the early warning of underwater weapons. Theinformation of shaft frequency and blade number of target can be attained through analyzingthe strong noise of line spectrum produced by propeller rotation. However, due to the strongambient noise in low frequency band in water, the method of analysis of DEMON spectrumwas used to extract line spectrum of target from demodulated information of target-radiatednoise in high frequency. Based on the extracted line spectrum, it was possible to obtain theshaft frequency of target and recognize the target. At the same time, the continuous spectrumpart of the ship radiated noise can also carry some information about the target. Whentransformed it with Hilbert-Huang Transformation, we will find out that each of its IntrinsicMode Function changes with different targets. According to this, we can classify ship targets.For the reason of the quantity and the complexity of the target feature that the Hilbert-HuangTransformation provided, it is difficult to figure out a result of the target classification with acertain formula. Then we can use the artificial neural network to process them because it isgood at summarization, association, analogy and spreading. In addition, the artificial neuralnetwork technology has high robustness and fault tolerance, so it has inherent advantages inprocessing the complex and changeable underwater acoustic signal.Recently, DSP technology is developing rapidly. As an embedded microprocessor, theDSP chip has a powerful data processing capabilities, as well as the support to the protocol ofthe network port, witch makes the signal processing system with a DSP core exchangeinformation with other devices conveniently. Compared to the PC platform, the DSP dataprocessing platform has smaller size, stronger ability to work independently and lower powerconsumption. So it is more suitable to use in practical applications. OMAP-L138is a floating,fixed-point compatible DSP produced by TI. It has the advantages of high-speed andlow-power consumption, and provide a good development platform for me to achieve theship target recognition algorithm on hardware.In this thesis, the analysis of DEMON spectrum, the algorithm of shaft frequencyextraction and the algorithm of extracting target features by Hilbert-Huang Transform wereall achieved on OMAP-L138platform. Eventually, a software system for target recognitionthat could obtain data from network port was established on hardware platform. In the end, some experiments were performed. We used DSP program to manipulate the experimentaldata and tested the performances of the two algorithms on hardware platform by these data. Itcould be seen from the statistical results that these two algorithms could both extract targetfeatures accurately and recognize targets correctly. At the same time, the algorithm of shaftfrequency extraction based on analyzing DEMON spectrum fulfilled the requirement ofreal-time. However, it was difficult to achieve real-time manipulation by using empiricalmode decomposition based on the algorithm of extracting target features by Hilbert-HuangTransform, because it was complicated and time-consuming. |
