本书的主要特点:
*新的特征生成技术,包括基于小波。小波包,分形的特征,还阐述了独立分量分析。
新增了关子支持向量机,变形模板匹配的章节,以及关于约束优化的附录。
特征选择技术。
线性以及非线性分类器的设计,包括贝叶斯分类器、多层感知器,决策树和RBF网络。
独立于上下文的分类,包括动态规划和隐马尔科夫建模技术。
不仅介绍了聚类算法的*新发展,而且还介绍了一些经典方法,诸如模糊。基因、退火等算法技术。
各种应用,包括图像分析。字符识别,医学诊断。语音识别以及信道均衡。

PrefaceCHAPTER1INTRODUCTION1.1IsPatternRecognitionImportant?1.2Features,FeatureVectors,andClassifiers1.3SupervisedVersusUnsupervisedPatternRecognition1.4OutlineoftheBookCHAPTERCLASSIFIERSBASEDONBAYESDECISIONTHEORY2.1Introduction2.2BayesDecisionTheory2.3DiscriminantFunctionsandDecisionSurfaces2.4BayesianClassificationforNormalDistributions2.5EstimationofUnknownProbabilityDensityFunctions2.5.1MaximumLikelihoodParameterEstimation2.5.2MaximumaPosterioriProbabilityEstimation2.5.3BayesianInference2.5.4MaximumEntropyEstimation2.5.5MixtureModels2.5.6NonparametricEstimation2.6TheNearestNeighborRuleCHAPTER3LINEARCLASSIFIERS3.1Introduction3.2LinearDiscriminantFunctionsandDecisionHyperplanes3.3ThePerceptronAlgorithm3.4LeastSquaresMethods3.4.1MeanSquareErrorEstimation3.4.2StochasticApproximationandtheLMSAlgorithm3.4.3SumofErrorSquaresEstimation3.5MeanSquareEstimationRevisited3.5.1MeanSquareErrorRegression3.5.2MSEEstimatesPosteriorClassProbabilities3.5.3TheBias-VarianceDilemma3.6SupportVectorMachines3.6.1SeparableClasses3.6.2NonseparableClassesCHAPTER4NONLINEARCLASSIFIERS4.1Introduction4.2TheXORProblem4.3TheTwo-LayerPerceptron4.3.1ClassificationCapabilitiesoftheTwo-LayerPerceptron4.4Three-LayerPerceptrons4.**lgorithmsBasedonExactClassificationoftheTrainingSet4.6TheBackpropagationAlgorithm4.7Variationsonthe;BackpropagationTheme4.8TheCostFunctionChoice4.9ChoiceoftheNetworkSize4.10ASimulationExample4.11NetworksWithWeightSharing4.12GeneralizedLinearClassifiers4.13Capacityofthe/-DimensionalSpaceinLinearDichotomies4.14PolynomialClassifiers4.15RadialBasisFunctionNetworks4.16UniversalApproximators4.17SupportVectorMachines:TheNonlinearCase4.18DecisionTrees4.18.1SetofQuestions4.18.2SplittingCriterion4.18.3Stop-SplittingRule4.18.4ClassAssignmentRule4.19DiscussionCHAPTER5FEATURESELECTION5.1Introduction5.2Preprocessing5.2.1OutlierRemoval5.2.2DataNormalization5.2.3MissingData5.3FeatureSelectionBasedonStatisticalHypothesisTesting5.3.1HypothesisTestingBasics5.3.2Applicationofthet-TestinFeatureSelection5.4TheReceiverOperatingCharacteristicsCROCCurve5.5ClassSeparabilityMeasures5.5.1Divergence5.5.2ChernoffBoundandBhattacharyyaDistance5.5.3ScatterMatrices5.6FeatureSubsetSelection5.6.1ScalarFeatureSelection5.6.2FeatureVectorSelection5.7OptimalFeatureGeneration5.8NeuralNetworksandFeatureGeneration/Selection5.9AHintontheVapnik--ChemovenkisLearningTheoryCHAPTER6FEATUREGENERATIONI:LINEARTRANSFORMS6.1Introduction6.2BasisVectorsandImages6.3TheKarhunen-LoeveTransform6.4TheSingularValueDecomposition6.5IndependentComponentAnalysis6.5.1ICABasedonSecond-andFourth-OrderCumulants6.5.2ICABasedonMutualInformation6.5.3AnICASimulationExample6.6TheDiscreteFourierTransform(DFT)6.6.1One-DimensionalDFT6.6.2Two-DimensionalDFT6.7TheDiscreteCosineandSineTransforms6.8TheHadamardTransform6.9TheHaarTransform6.10TheHaarExpansionRevisited6.11DiscreteTimeWaveletTransform(DTWT)6.12TheMultiresolutionInterpretation6.13WaveletPackets6.14ALookatTwo-DimensionalGeneralizations6.1**pplicationsCHAPTER7FEATUREGENERATIONII7.1Introduction7.2RegionalFeatures7.2.1FeaturesforTextureCharacterization7.2.2LocalLinearTransformsforTextureFeatureExtraction7.2.3Moments7.2.4ParametricModels7.3FeaturesforShapeandSizeCharacterization7.3.1FourierFeatures7.3.2ChainCodes7.3.3Moment-BasedFeatures7.3.4GeometricFeatures7.4AGlimpseatFractals7.4.1Self-SimilarityandFractalDimension7.4.2FractionalBrownianMotionCHAPTER8TEMPLATEMATCHING8.1Introduction8.2MeasuresBasedonOptimalPathSearchingTechniques8.2.1Bellman'sOptimalityPrincipleandDynamicProgramming8.2.2TheEditDistance8.2.3DynamicTimeWarpinginSpeechRecognition8.3MeasuresBasedonCorrelations8.4DeformableTemplateModelsCHAPTER9CONTEXT-DEPENDENTCLASSIFICATION9.1Introduction9.2TheBayesClassifier9.3MarkovChainModels9.4TheViterbiAlgorithm9.5ChannelEqualization9.6HiddenMarkovModels9.7TrainingMarkovModelsviaNeuralNetworks9.8AdiscussionofMarkovRandomFieldsCHAPTSR10SYSTEMEVALUATION10.1Introduction10.2ErrorCountingApproach10.3ExploitingtheFiniteSizeoftheDataSet10.4ACaseStudyFromMedicalImagingCHAPTER11CLUSTERING:BASICCONCEPTS11.1Introduction11.1.1ApplicationsofClusterAnalysis11.1.2TypesofFeatures11.1.3DefinitionsofClustering11.2ProximityMeasures11.2.1Definitions11.2.2ProximityMeasuresbetweenTwoPoints11.2.3ProximityFunctionsbetweenaPointandaSet11.2.4ProximityFunctionsbetweenTwoSetsCHAPTER12CLUSTERINGALGORITHMSI:SEQUENTIALALGORITHMS12.1Introduction12.1.1NumberofPossibleClusterings12.2CategoriesofClusteringAlgorithms12.3SequentialClusteringAlgorithms12.3.1EstimationoftheNumberofClusters12.4AModificationofBSAS12.**Two-ThresholdSequentialScheme12.6RefinementStages12.7NeuralNetworkImplementation12.7.1DescriptionoftheArchitecture12.7.2ImplementationoftheBSASAlgorithmCHAPTER13CLUSTERINGALGORITHMSII:HIERARCHICALALGORITHMS13.1Introduction13.2AgglomerativeAlgorithms13.2.1DefinitionofSomeUsefulQuantities13.2.2AgglomerativeAlgorithmsBasedonMatrixThetry13.2.3MonotonicityandCrossover13.2.4ImplementationalIssues13.2.**gglomerativeAlgorithmsBasedonGraphTheory13.2.6TiesintheProximityMatrix13.3TheCopheneticMatrix13.4DivisiveAlgorithms13.5ChoiceoftheBestNumberofClustersCHAPTER14CLUSTERINGALGORITHMSIII:SCHEMESBASEDONFUNCTIONOPTIMIZATION14.1Introduction14.2MixtureDecompositionSchemes14.2.1CompactandHyperellipsoidalClusters14.2.2AGeometricalInterpretation14.3FuzzyClusteringAlgorithms14.3.1PointRepresentatives14.3.2QuadricSurfacesasRepresentatives14.3.3HyperplaneRepresentatives14.3.4CombiningQuadricandHyperplaneRepresentatives14.3.**GeometricalInterpretation14.3.6ConvergenceAspectsoftheFuzzyClusteringAlgorithms14.3.7AlternatingClusterEstimation14.4PossibilisticClustering14.4.1TheMode-SeekingProperty14.4.2AnAlternativePossibilisticScheme14.5HardClusteringAlgorithms14.5.1TheIsodataork-Meansorc-MeansAlgorithm14.6VectorQuantizationCHAPTER15CLUSTERINGALGORITHMSIV15.1Introduction15.2ClusteringAlgorithmsBasedonGraphTheory15.2.1MinimumSpanningTreeAlgorithms15.2.2AlgorithmsBasedonRegionsofInfluence15.2.3AlgorithmsBasedonDirectedTrees15.3CompetitiveLearningAlgorithms15.3.1BasicCompetitiveLearningAlgorithm15.3.2LeakyLearningAlgorithm15.3.3ConscientiousCompetitiveLearningAlgorithms15.3.4CompetitiveLearning-LikeAlgorithmsAssociatedwithCostFunctions15.3.5Self-OrganizingMaps15.3.6SupervisedLearningVectorQuantization15.4BranchandBoundClusteringAlgorithms15.5BinaryMorphologyClusteringAlgorithms(BMCAs)15.5.1Discretization15.5.2MorphologicalOperations15.5.3DeterminationoftheClustersinaDiscreteBinarySet15.5.4AssignmentofFeatureVectorstoClusters15.5.5TheAlgorithmicScheme15.6BoundaryDetectionAlgorithms15.7Valley-SeekingClusteringAlgorithms15.8ClusteringViaCostOptimization(Revisited)15.8.1SimulatedAnnealing15.8.2DeterministicAnnealing15.9ClusteringUsingGeneticAlgorithms15.10OtherClusteringAlgorithmsCHAPTER16CLUSTERVALIDITY16.1Introduction16.2HypothesisTestingRevisited16.3HypothesisTestinginClusterValidity16.3.1ExternalCriteria16.3.2InternalCriteria16.4RelativeCriteria16.4.1HardClustering16.4.2FuzzyClustering16.5ValidityofIndividualClusters16.5.1ExternalCriteria16.5.2InternalCriteria16.6ClusteringTendency16.6.1TestsforSpatialRandomnessAppendixAHintsfromProbabilityandStatisticsAppendixBLinearAlgebraBasicsAppendixCCostFunctionOptimizationAppendixDBasicDefinitionsfromLinearSystemsTheoryIndex

模式识别在所有的自动化,信息处理和检索应用中都至关重要。本书由该领域内的两位**专家合著而成,从工 程角度,全面阐述了模式识别的应用,涉及的主题从图像分析到语音识别与通信,书中涉及到了神经网络的前沿材料, 着重描述了包括独立分量和支持向量机在内的*新进展。本书是享誉世界的名著,经过十余年的发展,已成为此领域 *全面的参考书,被世界众多高校选用为教材。除了适合教学外,也可供工程技术人员参考。
本书的主要特点:
*新的特征生成技术,包括基于小波。小波包,分形的特征,还阐述了独立分量分析。
新增了关子支持向量机,变形模板匹配的章节,以及关于约束优化的附录。
特征选择技术。
线性以及非线性分类器的设计,包括贝叶斯分类器、多层感知器,决策树和RBF网络。
独立于上下文的分类,包括动态规划和隐马尔科夫建模技术。
不仅介绍了聚类算法的*新发展,而且还介绍了一些经典方法,诸如模糊。基因、退火等算法技术。
各种应用,包括图像分析。字符识别,医学诊断。语音识别以及信道均衡。
作者简介:
Sergios Theodoridis 是希腊雅典大学信息系教授。于1973年在雅典大学获得物理学学士学位,又分别于 1975年,1978年在英国伯明翰大学获得信号处理与通信硕士和博士学位。主要研究方向是自适应信号处理。通信与模式识别。他是欧洲并行结构及语言协会(PARLE-95)的主席和欧洲信号处理协会(亡USIPCO-98)的常务主席、《信 号处理》杂志编委。
Konstantinos Koutroumbas 任职于希腊雅典**天文台空间应用研究院,是国际知名的专家。