Active Phased Array Radar Signal Design And Lpi Performance Research

With the diversification of inter personal communication, the communicationbetween different languages also appears more and more frequent. Languagerecognition technology as a branch of speech signal processing technology has beenpain more and more attention and research in recent years. The populartext-independent language recognition technology mainly based on acoustic model andphoneme sequence model, then established the matching recognition model whichmainly include PRLM, GMM, SVM. In order to reduce the dimension of thecharacteristic parameters, this thesis studies mainly focus on trying to introduce themanifold learning algorithm to the study of language recognition, and use supportvector machine (SVM) for comparing with the based recognition model.The main work and contribution of this thesis are outlined as follows:The first research work of this paper is mainly on analysis of language phoneticfeatures, aiming at the characteristic between different languages, to extract thecharacteristic parameter vector which represents language information.The second research work of this paper is mainly on the global optimal dimensionreduction algorithm of the principal component analysis (PCA) applied in the field oflanguage recognition. Through deeply research in PCA, we proposed the PCA to solvethe problem of large training language models in language recognition systems.Through the experiments, this paper has proved that the feasibility of the application ofprincipal component analysis (PCA) to language identification technology.The third research work of this paper is mainly on the application of the LPPalgorithm in the language recognition technology. On the basis of the algorithm,considering the local information not describe data and the defect of the algorithm doesnot take into account the category information, the LAC-LPP algorithm based on theattribute of the constraints of local protection distance projection was raised. With theclose relevance of package, algorithm based on the characteristic parameters in thedistribution of space according to the concentration inside the class, class room awayfrom the principle, designed feature parameter local geometrical structure of data points. In the experiments, applied the three kinds of manifold learning algorithms to theextracted speech feature parameters for feature dimension reduction processing, thenchoose the nearest neighbor classifier for classification and recognition, goodclassification effect of the proposed algorithm was proved.