1 Introduction
- 格式:pdf
- 大小:217.76 KB
- 文档页数:4
主要内容 (Outline)• 绪论小规模集成电路三(SSI)• 逻辑函数基础 门电路个• 组合逻辑电路模 块中规模集成电路 (MSI)• 集成触发器 • 时序逻辑电路大规模集成电路 • 半导体存储器(LSI)• 数模、模数转换电路绪论 (Introduction)一、数字(digital)信号和模拟(analog)信号 数字量和模拟量 数字电路和模拟电路二、数字信号相关概念 二进制数 Binary Digits 数字信号的逻辑电平 Logic Levels 数字信号波形 Digital Waveforms一、Digital Signal and Analog Signal Digital and Analog Quantities电子 电路 中的 信号模拟信号: 连续analogue signal value数字信号: 离散digital signal valuetime time模拟信号T( C) 30采样信号T( C)sampled3025离散化 2520202 4 6 8 10 12 2 4 6 8 10 12 t (h)A.M.P.M.2 4 6 8 10 12 2 4 6 8 10 12 t (h)A.M.P.M.数字化-表示 为由0、1组成 的二进制码Analog Electronic SystemDigital and Analog Electronic System★ 工作在模拟信号下的电子电路是模拟电路。
研究模拟电路时,注重电路输入、输出信号 间的大小、相位关系。
包括交直流放大器、 滤波器、信号发生器等。
★ 模拟电路中,晶体管一般工作在放大状态。
★ 工作在数字信号下的电子电路是数字电路。
研究数字电路时,注重电路输出、输入间的逻 辑关系。
主要的分析工具是逻辑代数,电路的 功能用真值表、逻辑表达式或波形图表示。
★ 在数字电路中,三极管工作在开关状态, 即工作在饱和状态或截止状态。
Unit 1 Introduction Text AKey to the ExercisesReading ComprehensionI. read text A and answer the following questions.1. Fair hair and blue eyes.2. Mouth, nose, expression, and much of her character.3. Because her home is very important to her.4. Making models, decorations and candles.5. Because she finds homework is boring.II.Choose the best answers according to Text A.1-5 DCADDV ocabulary ExercisesI. Fill in the blanks with correct form of the words and phrases given below.1. temper2. make friends3.On the whole4.as well as5.keen6.behaves7.tidy8.argument9.boring 10.violent II.Choose the correct form in the bracket to complete each sentence.1.bored2. bad-tempered3.final4.behavior5.argue6.decorate7.violently8.abroadIII. Use the words or expressions you have learned in the text to replace the following words in italics.1.On the whole2.an argument3.are behaving4.boring5.make friends6. keen on7.tidy8.violentStructure ExercisesI. Now make similar sentences with the words and expressions given below.1. Most people would rather stay home.2. I would rather not talk about it.3. She would rather walk home after work.4. I would rather go to the beach this weekend.II.Now join the following pairs of sentences by using “because”.1. She’s studying because she has a test tomorrow.2. John didn’t attend the meeting because he was ill.3. We did n’t enjoy the day because the weather was so bad.4. I decided to go with them because I had nothing else to do.5. She’s in a bad mood because her father doesn’t allow her to see her boyfriend tonight.Translation Exercise1.The street looks like a garden.2.Her daughter wants to study abroad for a year.3.the store sells newspapers, magazines as well as picture books.4.I would rather spend the weekend in the countryside.5.Our manager is away on holiday this week.Text BKey to the ExercisesReading ComprehensionChoose the best answer according to Text B1.C2.C3.D4.A5.AV ocabulary ExercisesI. Fill in the blanks with correct form of the words and phrases given below.1. Maybe2. pleasure3. would like4. have got to5.yet6. associates7.settled8.represent9.pleased 10.introduce.IV. choose the most suitable answer for each of the following sentences.1-5 ADABC 6-10 DADBCTranslation Exercise1.我觉得他想现在回家。
1 Introduction教学目的: List six different property classifications of materials that determine their applicability. Cite the four components that are involved in the design, production and utilization of materials, and briefly describe the interrelationship between these components.教学重点: The four components that are involved in the design, production and utilization of materials教学难点: The discipline of materials science involves investigating the relationships that exist between the structure and properties of materials.教学方法:Multimedia学时分配1.1Historical Perspective10 min1.2Materials science and engineering 25 min1.3Why Study Materials Science and Engineering 10 min1.4Classification of Materials 35 min1.5 Modern Material‟s Needs 10 min教学过程及主要内容:1. Historical PerspectiveWebster编者“New International Dictionary(1971年)”中关于材料(Materials)的定义为:材料是指用来制造某些有形物体(如:机械、工具、建材、织物等的整体或部分)的基本物质(如金属、木料、塑料、纤维等)迈尔《新百科全书》中材料的含义:材料是从原材料中取得的,为生产半成品、工件、部件和成品的初始物料,如金属、石块、木料、皮革、塑料、纸、天然纤维和化学纤维等等。
高教版中职英语基础模块1全套教案教案一:Unit 1 Introduction to English教学目标:1. 了解英语的起源和发展历程。
2. 学习英语的基本发音规则和语音特点。
3. 掌握英语的基本问候语和自我介绍的表达方式。
教学重点:1. 英语的起源和发展历程。
2. 英语的基本发音规则和语音特点。
3. 英语的基本问候语和自我介绍的表达方式。
教学难点:1. 英语的发音规则和语音特点的掌握。
2. 自然流利地运用英语的问候语和自我介绍。
教学准备:1. 多媒体设备。
2. 讲义和练习题。
教学过程:Step 1: Lead-in1. Greet the students and introduce the topic of the lesson.2. Show pictures of different countries and ask students if they know any English-speaking countries.3. Ask students why they think English is important.Step 2: Presentation1. Introduce the origin and development of English.2. Show a timeline of the major events in English history.3. Explain the basic pronunciation rules and phoneticfeatures of English.4. Use audio or video materials to demonstrate the correct pronunciation.Step 3: Practice1. Divide the class into pairs or small groups.2. Give students a list of common greetings and ask them to practice using them in different situations.3. Have students practice introducing themselves to each other using the phrases and sentences learned.Step 4: Consolidation1. Review the key points of the lesson, including theorigin and development of English, pronunciation rules, and greetings.2. Ask students to summarize what they have learned intheir own words.Step 5: Assessment1. Give students a short quiz to test their understanding of the lesson.2. Assign homework, such as writing a short paragraph about the importance of English or practicing greetings and self-introductions.教案二:Unit 2 Numbers and Time教学目标:1. 学习基本的数字和时间表达方式。
HLP ImplementationVersion0.1International Computer Science Institute1IntroductionHybrid Link-State Path-Vector Protocol,or HLP,is an inter-domain routing protocol designed as a replacement for the current Border Gateway Protocol(BGP).Using a combination of link-state and path vector routing,it provides greater scalability,better fault isolation and better convergence.The core of HLP is the inclusion of the economic and political structure of the Internet into inter-domain routing.That is,BGP currently considers each AS as a node in a general graph without any specific structure(using explicit policies to constrain routing),whereas HLP assumes that the Internet structure is basically hierarchical with the provider autonomous systems(ASs)being the roots of customer ASs.HLP explicitly includes the relationship between2neighboring ASs in its protocol.This will reduce misconfigurations which should hopefully reduce the occurrence of routing abnormalities.However,the tradeoff is some amount of inflexi-bility in the routing algorithm.This is resolved using exceptions that are expected to be rare and therefore acceptable. This report summarizes implementation of HLP on the XORP[1]software router.The implementation reuses much of the code in XORP’s BGP module.2DefinitionsWe say that two ASs are in the same hierarchy if there exists a directed path between them such that the path consists of any number of provider links followed by any number of customer links.This definition of hierarchy implies that there exists at least one route between two ASs in a hierarchy that does not include(provider)(customer)(provider)links. Two ASs are neighbors if there exists a link between them.The relationship between neighboring ASs(peer,customer, or provider)determine the overall structure of the network,which can be as simple as shown in Figure1a,or consist of overlapping hierarchies as shown in Figure1b.In thefigure,each node represents an AS;the neighboring AS at a higher tier is the provider,similarly an AS at a lower tier is the customer.Thus,AS A is the provider of AS B,which in turn is A’s customer.Neighboring ASs at the same tier are also called peering ASs.Note that the use of tiers in Figure1and inprotocol.subsequentfigures only allows for graphical representation of relationships,it is not present in the actual HLPHLP divides the network into hierarchies consisting of providers and their customer ASs,and peering ASs in different hierarchies allow routing between hierarchies.As will be explained in the next section,this division of the network intoseparate components increases scalability,as well as reduces the convergence time for route updates.3Routing Information DisseminationTwo types of routing packets are used to disseminate routing information:link-state advertisements(LSAs)and fragmented path-vector(FPV)1.As is the case in OSPF,LSAs areflooded throughout a hierarchy,and allow construction of the entire hierarchy topology.FPVs are used to route between ASs;they contain the numbers of peering ASs between hierarchies. LSAs that have not previously been received are forwarded in the following manner:1.if they are from customers,we forward to all neighbors,except the customers from which they arrive.2.if they are from providers,we forward only to neighboring customers,not providers.The objective of the above rules is to restrict LSAs to the hierarchies from which they originate.Failure to do so will imply that multi-homing ASs belonging to different hierarchies can cause LSAs to be disseminated throughout the entire Internet. The rules above implements this restriction,illustrated in Figure2.By definition,AS C belongs to both hierarchies1and 2,and rule1allows LSAs from AS B to be forwarded to C.At C,rule2prevents LSAs from A from being forwarded to AS D.On the other hand,LSAs involving C will be disseminated in both hierarchies,which is correct since C is a member both.ofFPVs are used to forward routes from one hierarchy to another.They are similar to path-vector packets used in BGP,except that they do not include the AS path within hierarchies.Rules that govern forwarding of FPVs are given as follows:1.FPVs from providers are disseminated only to customers,not to peers or other providers.2.FPVs from peers are forwarded to neighboring peers and customers,but not providers.4Routing AlgorithmThe mechanism to choose a route to a particular destination AS is similar to that currently used in BGP.This eases the implementation of exceptions which will be covered in the next section,as well as the creation of FPVs for routes to customers within the hierarchy Basically,we store routes for destination ASs reachable from each neighboring AS,then decide the winning route for a particular destination.The handling of routes contained within FPVs is straightforward and similar to that of BGP,but routing information gained from LSAs needs to be converted to the same form as that in FPVs. The conversion is elaborated on in the next section.4.1From LSAs to RoutesWe denote the least cost of reaching AS A from B by cost(B,A),a route from A to B with cost C by route[(A,B),C], and we perform the conversion in the following manner,assuming that the operations are taking place in AS X:1.for each neighbor N2.if neighbor is a customer3.for each downstream customer AS Apute cost(N,A)5.create AS path[X,A]with cost C=[cost(N,A)+cost(X,N)]6.associate route[(X,A),C]with N7.else if neighbor is a provider8.for each of the non-customer ASs in the hierarchypute cost(N,A)10.create AS path[X,A]with cost C=[cost(N,A)+cost(X,N)]11.associate route[(X,A),C]with NThe computation is broken into two parts,steps2to5,and6to9of the algorithm above.This is required because forwarding of routes from a provider to a peer should take place only if the destination AS is a customer2.To distinguish between the origin of the routes,we tag them with the following:PROVIDERLSA:Route for non-customer destination AS within the same hierarchy,determined using link-state information.PEERLSA:Route for customer destination AS obtained using link-state information.An example is given in Figure3,where we focus on AS C.Table1shows the routes,costs and tags associated with each neighbor.Neighbor Cost(C,A)PROVIDERA3LSA(C,E)PROVIDERA15LSA(C,D)PROVIDERD26LSA(C,A)PROVIDERF5LSATable1:Table containing routes,costs and tags for example converting LSAs to routes4.2From FPV to RoutesCreation of routes from FPVs are much simpler,and is similar to that in BGP:1.if FPV is from a provider P2.extract route and metric,tag with PROVIDER_FPV,and associate them with P3.if FPV is from a peer Q4.extract route(prepending this AS’number)and metric,tag with PEER_FPV,and associate with Q5.if FPV is from customer,treat FPV as coming from peer,extract route(prepending this AS’number)and metric,tag with PEER_FPV,and associate with QNote that step5only occurs due to an exception in the customer AS.Figure3:Example illustrating LSA to route conversion4.3Route SelectionThe winning route to a particular destination is selected according to the following order of preferences:1.customer route(ie.tagged with CUSTOMERNeighbor TypePROVIDER CustomerFPVCUSTOMER PeerTable2:Route types that can be forwarded to corresponding neighboring AS types5ExceptionsHLP supports three different types of exceptions.The primary use of exceptions is to support operations that are currently used in BGP but not covered by the default HLP rules.The format in which exceptions are specified and stored is given by wherefromlink specifies the neighboring link to which winning routes will be propagated,andAS number refers to the destination AS that this particular exception is for.In HLP,a from link is given by the tuple:5.1Exception1Figure4:(a)Example illustrating effect of exception1on rest of hierarchy.D announces that it does not have a customer route to C.B uses graph in(b)to compute shortest paths to all ASs in the same hierarchy except C(i.e.A,D and F).B uses the graph in(c)to compute the shortest path to C.Thefirst exception,illustrated in Figure4,allows an AS(say D)to choose an alternate route to a customer(C)via a peer (E).The route choose is dependent on the customer AS,and should not affect routes to other ASs.D indicates its intention via LSAs to other ASs in the same hierarchy that it is not choosing customer routes to C.D also informs E that it no longer has a customer route to C.If,ignoring customer routes,the winning route is from E,then D forwards that FPV to its peers and customers.However,if the winning route is from another peer not specified in the exception,then the corresponding FPV will not be forwarded.Using the same example,exception1is specified bywhere is specified using the tuple5.2Exception2This exception specifies that winning routes from the stated provider is to be forwarded to a particular peer,which is typically not done.In Figure5,the exceptionallows winning routes to AS Z from provider A to be forwarded to D.5.3Exception3The last exception supported is similar to exception2.Here,routes are forwarded from a peer to a provider.Currently, the provider simply accepts incoming FPVs from customers,treating them as though they are from peers.If,for security reasons,providers should reject FPVs from customers,then additional configuration will be required in the provider.Figure5:Simple network illustrating route forwarding from provider to peer5.4Exception Format and MatchingThe type of exception does not explicitly need to be specified.Instead,the neighbor relationship associated with the links given in each exception rule can be used to determine this.The format is thus simplified,and should hopefully reduce the occurrence of misconfigurations.Table3gives the combination of links that indicate the kind of exception specified.From To Exception TypeNULL1PeerPeer3Table3:Combination of link types associated with each exception type6Data StructuresIn this section we describe the data structures used to maintain state in a HLP router.Figure6shows theflow of routing information through the system,as well as the major components of the system.Figure6:Flow of routing information through systemWe describe each component below:Peer:A peer contains the necessary objects needed for receiving and sending routing information from and to neighboring routers.Each peer maintains the state machine for the connection associated with the corresponding neighbor,as well as the various timers needed during connection establishment and for keepalive messages.RibIpcHandler:Handles insertion of prefixes owned by this AS.Routing Information Base Input Table(RibInTable):Stores routes associated with corresponding neighboring router.Routes may be obtained via FPVs sent from neighbor,or from computation of shortest paths using link-state infor-mation.The type of tag assigned to a route is explained in Section4.1.PeerHandler:Handles reception and transmission of FPVs and LSAs.After a packet has been received,it is broken up into individual components(for instance,link changes,route withdrawals,route announcements,etc.)before being passed to HLPCore for processing.Updates passed from the DecisionTable are aggregated as much as possible within a packet(FPV or LSA)before being transmitted.Routing Information Base Output Table(RibOutTable):Stores the outgoing routes sent via FPVs.The contents of this table is a subset of the corresponding RibInTable of the neighboring router.PeerData:Contains information related to the peering link:–neighbor’s AS number and identification number(ID),–IP tuple of the connection,–neighbor’s peer type(customer,provider or peer),–metric,or cost of the link–various timeout values(hold,retry,keepalive)HLPCore:The HLPCore contains the Lib,DecisionTable,ExceptionTable,and the AS-Prefix Map.It maintains the periodic update timer3,and interfaces between the user and the system.The core also connects Peers with the Lib and DecisionTable,so that incoming routing information can be processed and then pushed out of the system if necessary.Link-State Information Base(LIB):The Lib stores link-state information gathered from LSAs received.The network topology constructed is then used to determine the shortest path to each destination AS from every neighbor.DecisionTable:The DecisionTable chooses the winning route amongst the routes stored in the RibInTables for a particular destination prefix.The selection is based on the order of preferences given in Section4.3.Winning routes are stored in a trie,after which they are pushed to the neighboring routers based on the route type as specified in Section4.4.In general the DecisionTable deals with FPVs,whilst the Lib deals with LSAs.ExceptionTable:The ExceptionTable stores the exceptions raised locally,as well as those raised by other ASs within the same hierarchy(exceptions raised are not explicitly propagated to other hierarchies).Currently,only information with regards to exception1are disseminated via LSAs.The ExceptionTable is used when the DecisionTable is determining whether a particular winning route should be sent to a neighbor,and when the Lib is computing the shortest path to destination ASs taking into account exception1s raised.AS-Prefix Map:This object stores the mapping of ASs to the corresponding advertised prefixes.Since the core of HLP manages routes at the prefix level4,and route changes are disseminated at the AS level,the AS-Prefix Map is required to translate between the two.Thus,for instance,incoming route changes(which will not include the prefixes involved,but just the AS path)will be processed in the HLPCore at the prefix level,and merged again just before the updates are pushed out.7Finite State MachineThefinite state machine for each peering connection is the same as that specified in[2].8Protocol FormatThe message header format as specified in [2]remains unchanged,as is the format of the Keepalive packet.In this section we describe the changes to the Open and Update (renamed Fragmented Path-Vector)packets,as well as introduce the LSA packet.8.1Open PacketVersionMy AS numberHold timeHLP identifierPeer type12241Figure 7:Format of Open packet,numbers denote lengths of corresponding fields in octetsSince a HLP network is dependent on the relationship between neighboring ASs,it is important that they agree on that.We thus include an additional field in the Open packet,the peer type field.The relationship type inserted in the field is with respect to the neighbor.For instance,if AS A is a customer of B,it will insert type corresponding to customer in the field of the Open packet it sends to B.Inconsistencies will result in the connection failing.8.2Fragmented Path Vector PacketThe format of the FPV ,shown in Figure 8,is similar to BGP’s Update packet,but with an additional AS Down field.When an AS becomes unreachable for any reason,rather than withdrawing every route to that AS,we propagate just the AS number in this field.Unfeasible routes lengthWithdrawn routesAS down lengthASnumbersTotal path attribute lengthPath attributesNetwork layerreachability informationX X2X 2X 2Figure 8:Format of FPV packet,the numbers denote lengths of corresponding fields in octets.An X means that the field length is variable.The format in which a network prefix is represented is shown in Figure 9.This representation is used for the network layer reachability information (NLRI)and withdrawn routes in Figures 8and 10.Length1PrefixX Figure 9:Representation of a prefix:the Length field uses 1octet,and the size of the Prefix field is variable.8.3Link State Advertisement PacketThe LSA packet is a new packet type,and the fields are shown in Figure 10.We describe the four main fields and their corresponding subfields as follows:1.link changes :This major field contains information regarding links grouped together according to an endpoint AS’number.For instance,all links stated in “link information (1)”have an end AS with number “AS (1)number”.Thus,multiple link changes can be aggregated and transmitted within the same packet.The format in which information for each link is transmitted is shown in Figure 11.2.unreachable ASs :These fields provide the list of ASs (“Unreachable ASs”)that are declared unreachable from an AS (“AS unreachable from”).This field is used to disseminate an AS’setting of exception 1.lengthLink changesAS unreachable from (1)UnreachableAS length (1)UnreachableASs (1)Unreachable AS lengthReachable AS length AS reachablefrom (1)ReachableAS length (1)ReachableASs (1)AS reachablefrom (n)ReachableAS length (n)ReachableASs (n)Unfeasible routes length Withdrawn routesNLRI total length AS announcingNLRI (1)NLRIlength (1)AS (1)numberLink changeslength (1)Linkinformation (1)NLRI (1)AS announcingNLRI (n)NLRIlength (n)AS unreachablefrom (n)UnreachableAS length (n)UnreachableASs (n)AS (n)numberLink changeslength (n)Linkinformation (n)222 222X2222NLRI (n)2222X2X22X22XXX X22XFigure10:Format of LSA packet.The numbers representfield size in octets,X means thefield size is variable.Neighbor AS number NeighborrelationOperation Reserved Metric1622432Figure11:Link information representation format.The numbers representfield size in bits.3.reachable ASs:Thesefields provide the list of ASs(“Reachable ASs”)that are declared reachable from an AS(“ASreachable from”).Note that the list of ASs must have previously been declared unreachable.Thisfield is used when deleting exception1.4.unfeasible routes:Similar to Update packets,thisfield holds the routes that have been withdrawn by ASs in the samehierarchy.5.NLRI announcements:Thefinalfield gives the prefixes that each AS is announcing.9Boot Up ProcedureThe HLP protocol does not require knowledge of the tier level an AS is at.When connection to a new neighbor is established,the following steps are taken:if neighbor is a peersend all routes tagged with PEER_FPV and CUSTOMER_LSAif neighbor is a customersend all routes tagged with PEER_FPV and CUSTOMER_LSAsend all link-state informationsend all exception informationif neighbor is a providersend all link-state information for customer ASssend all exception informationAdditional routes that match exceptions,if any,are also sent.10Exception Setting and RemovalHLP allows dynamic setting and deletion of exceptions.Setting of exceptions should cause the network state to become the same as if the exceptions were present on bootup.Similarly,deletion of exceptions should cause the state to be the same as if the exceptions were never present.The following steps are taken when the corresponding exceptions are raised or removed:Exception1:Upon setting of exception1,the AS broadcasts an LSA packet in its hierarchy indicating the lack ofa customer route to the destination AS.It removes route(s)to the destination AS(tagged with CUSTOMERLSA and PROVIDER。
一、课题:《全新版大学英语1》Unit 1 Introduction二、教学目的:1. 帮助学生掌握英语基本句型和常用词汇;2. 培养学生的英语听、说、读、写能力;3. 激发学生学习英语的兴趣,提高学生的英语综合素质。
三、课型:新授课四、课时:2课时五、教学重点:1. 掌握英语基本句型和常用词汇;2. 培养学生的英语听、说、读、写能力;3. 理解文章主旨,提高阅读理解能力。
六、教学难点:1. 英语基本句型的运用;2. 阅读理解中的长难句解析;3. 英语写作技巧的掌握。
七、教学过程:(一)导入新课1. 教师播放一段与英语学习相关的视频,激发学生的学习兴趣;2. 提问:视频中涉及哪些英语学习技巧?如何将这些技巧运用到实际学习中?(二)讲授新课1. 教师讲解英语基本句型,如:What's your name? How old are you? 等;2. 教师带领学生进行句型练习,巩固所学知识;3. 教师讲解常用词汇,如:name、age、school、home 等;4. 教师引导学生进行词汇练习,提高词汇运用能力;5. 教师带领学生阅读课文,讲解文章主旨和段落大意;6. 教师解析阅读理解中的长难句,帮助学生提高阅读理解能力;7. 教师讲解英语写作技巧,如:如何组织文章结构、如何运用过渡词等。
(三)巩固练习1. 教师组织学生进行句型练习,巩固所学知识;2. 教师组织学生进行词汇练习,提高词汇运用能力;3. 教师组织学生进行阅读理解练习,提高阅读理解能力;4. 教师组织学生进行英语写作练习,提高写作技巧。
(四)归纳小结1. 教师总结本节课所学内容,强调重点和难点;2. 教师提醒学生在课后进行复习,巩固所学知识。
(五)作业安排1. 复习本节课所学英语基本句型和常用词汇;2. 阅读课文,完成课后练习;3. 按照所学写作技巧,写一篇英语短文。
八、板书设计:全新版大学英语1 Unit 1 Introduction一、英语基本句型:1. What's your name?2. How old are you?3. Where do you come from?4. What do you do?二、常用词汇:1. name2. age3. school4. home三、阅读理解:1. 理解文章主旨;2. 解析长难句;3. 提高阅读理解能力。
Chapter 1Introduction1. Define the following terms briefly.(1) linguistics语言学: the scientific or systematic study of language.(2) language语言: a system of arbitrary vocal 任意的声音symbols used for human communication.用于人类交流的任意声音符号系统(3) arbitrariness任意性: the absence of similarity betweenthe form of a linguistic sign and what it relates to in reality,语言符号的形式与现实的关系缺乏相似性e.g. the worddog does not look like a dog.(4) duality双重性: the way meaningless elements of languageat one level (sounds and letters) combine to formmeaningful units (words) at another level.在一个层面上(语言和字母)的无意义的语言元素结合在另一个层次上形成有意义的单位(词)(5) competence语言能力: knowledge of the grammar of alanguage as a formal abstraction and distinct from thebehavior of actual language use作为一种形式抽象的语言的语法知识,区别于实际语言使用的行为, i.e.performance.(6) performance语言运用: Chomsky’s term for actuallanguage behavior as distinct from the knowledge thatunderlies it, or competence.乔姆斯基对实际语言行为的术语不同于它的知识,或能力。
Unit 1 introduction一.文化文化是冻结了的人际交流,而交流是流动着的文化----W.B. Pearce, 1994.背景:长期以来,文化被认为是无处不在,无所不包的人类知识和行为的总体。
被笼统地当作“生活方式”,社会生活的一切方面,积淀物,价值观念体系,众多规范,乃至艺术,政治,经济,教育,修养,文学,语言,思维的总和。
概括地讲,文化即是人们所思,所言,所为,所觉的总和。
在不同的生态或自然环境下,不同的民族创造了自己特有的文化,也被自己的文化所塑造。
It is said that there are at least 150 definitions about culture.“Culture may be defined as what a society does and thinks”(Sapir, 1921) “Culture is man’s medium, there is not one aspect of human life that isnot touched and altered by culture. This means personality, how people expressthemselves, including shows of emotion, the way they think, how they move, howproblems are solved, how their cities are planned and laid out, how transportation systems function and are organized, as well as how economic and government systems are put together and fuction.” (E.T. Hall,1959)“A culture is a collection of beliefs, habits, living patterns, andbehaviors which are held more or less in common by people who occupy particular geographic areas” (D.Brown, 1978)文化的特性:1). 文化是由人们的内隐和外显行为组成的。
Shaping with PatternsJiayuan Meng Department of Computer Science University of Virginiajm6dg@December14,20061IntroductionPeople like magic.It is fun to imagine a cloud is shaped like a Mickey mouse,or the pattern of leaves andflow-ers are associated with human face.Sometimes infiction movie or cartoon production,we like to see similar visual effects.In this paper,we define the problem as following: Given two images,one for pattern and one for shape.We output another image,which draws the shape provided in the shape image,but using the patterns defined in the pat-tern image.This is a typical texture synthesize problem with user interaction.In a typical texture synthesize prob-lem,a source pattern is used to produce another texture, which visually resembles the pattern.Some algorithms have also been invented to guide the texture synthesize process by providing another user defined image as a hint to grow the texture.Based on these algorithms,we suc-cessfully solved the problem stated above.2Related WorkTexture synthesize has been widely discussed.Efros and Leung[1]used markov randomfields to grow a new tex-ture from a seed cut from the original texture.Hertzmann et al.has generalized the idea of MRF and the multi-resolution method,the resulting technique is image anal-ogis[2].The program learns the features from a pair of image,A and A’.A’is thefiltered image of A.Given a new image B,the program can produce image B’with the sim-ilarfilter effect.Image analogies is a powerful tool and it can achieve multiple effects including texture synthesize, texture transfer,and texture-by-numbers.Our algorithm is mainly based on image analogies.3Methodology3.1Brief Review of Image AnalogiesImage analogies works briefly as follows:1.Build the gaussian pyramid for the source image pairA and A’,as well as the target unfiltered image B.2.From the coarser level to thefinest level,build the gaussian pyramid of thefiltered image B’.3.At each level,we construct image B’row by row. At each pixel p,we search for a corresponding position s(p)in A’in the same level.Then we compute B’(p) from the feature at s(p)in A’.The computed pixel is an optimization of an energy function which attempts to maintain both the approximation and the coherence.We describe each step in a bit more detail:First,a user has to define a feature for each pixel.The feature is selected by the ually it is simply the RGB color or the luminance.Note that the feature for image A can be different from that of image A’.However, the feature of A and B are usually the same,and so as the features of A’and B’.Next,we need to construct the feature vector.For the source image pair A and A’,we build a gaussian pyramid.A feature vector F(i,j)is associated with a position(i,j), and it is a concatenation of features in the5*5neighbor-hood of the current level and the3*3neighborhood of the coarser level on both the images.So in this case,the feature vector for the source image pair has a length of68.We compute the feature vector for the image pair B and B’in the same way.The only difference is that when B’is constructed row by row,we don’t know all the feature around the current pixels neighborhood.Therefore,we neglect those invalid neighborhood pixels and build a partial feature vector.We use F a to denote the feature 1vector for image pair(A,A’),and F b to denote the feature vector for image pair(B,B’).Now we have defined the feature vectors.When looking for the best position s(p)for the current pixel in B’,wefirst do an ANN search tofind the approximate nearest neighbor in the feature vector space of the image pair(A,A’)at the current level.Care must be taken since F a has a length different with F b.We resolve this by selecting in F a only thefields that also present in F b.We denote this feature vector F appIn addition tofinding the feature in F a which best ap-proximates F b,we also have tofind the F a that exhibits the best coherence.We search in the neighborhood of p in B’.Each resolved neighbor q is associated with a source position s(q)in A’.Suppose p is an offset of r from q(p=q+r),we thenfind the F a(s(q)+r)that best approximates F b(p).we denote this feature vector F coh.The user has to define a coefficient K to select between F app and F coh.Basically,if F app<K∗F coh,then we select F coh,otherwise,F app is selected.We assign s(p)to the the position in A’that corresponds to the selected feature vector.3.2Analysis of Image AnalogiesImage analogies has two functions that are similar to our goal of rendering a shape with apattern.Thefirst is texture transfer,illustrated in Figure1.It uses the same fabric pattern image for A and A’,and set the target image to be B.Thefiltered image is a girl’s face rendered with the fabric texture.However,the pattern is applied globally.Although e can confine thefilter to be applied to certain boundaries,that will result in hard boundary edges that doesn’t”bleed”and merge naturally with the background.Anotherfilter is texture-by-number.In which A will be a segmentation of A’denoted with numbers(or different colors).By providing another segmentation image with the same set of numbers,we can construct a B’which has heterogeneous textures arranged according to the seg-mentation.This is illustrated in Figure2.This turns out to be very similar to our goal.The only difference is that we need to render a foreground with the pattern,and the texture-by-number is rendering a numbered image with the pattern.3.3Extended Texture-by-numberGiven the above observation,we made an attempt to build a numbering of the pattern image from the target image.First,we have to define the numbering of the pattern image.Illuminance is a good candidate.However,in most conditions,we want to distinguish the foreground from the background,so that the target shape can be rendered using only the foreground pattern.Therefore, segmentation information is required(Figure4).The resulting numbering is two fold:a segmentation mask, Figure1:Texture transfer using image analogiesFigure2:Texture-by-number using image analogies2and the illuminance of the foreground.Thus,A is a gray scale image with a matt.Second,we have to convert the target shape in to the numbering defined above.We follow the same procedure.First,we encode the segmentation information of the target image in a mask,then we add the illuminance of the foreground.B is also a gray scale image with a mask.To provide better results,the illuminance of B is nor-malized to the illuminance of A.Histogram match is an-other way which provides similar results.We can then define the feature of A and B to be a vector of 2values [α,l],where αis the matt value,and l is the illuminance.Now we have the input images ready,we startimage analogies to produce B’.This is illustrated in Figure 3.Figure 3:Raw result by numbering texture with segmentation and illuminationFigure 4:matting is important,without matting,the foreground pattern will be confused with the background pattern 3.4Key Point EnhancementAlthough we have preserved shape,there are still someimportant features missing.Some positions in B is more detailed and is the key feature for B,such as the eyes of a face,etc.At these key points,we want B’to resemble B more than anywhere else.We solve this by first finding the key points in B using the SIFT method by Lowe [3].In general,SIFT works by finding the positions where the image has a local extrema in the Laplacian of Gaussian field.If the image has significant contrast at this point,then we regard it as a key point.Key points are found at every level of the Gaussian pyramid.At each level,we decrease the value of K at each key point so that B’can preserve more similarity with B at that point.On the other hand,coherence is enhanced at places where there is no key point and the image exhibits low contrast.The resulting image is shown in Figure 54ConclusionWe have also tried synthesize in B’s gradient field,and then use poisson editing [4]to recover the target shapeFigure 5:Enhanced result by preserving more details at key points.The similarity with B can be further tuned globally byuser.3with a gradient style of the given patterns.However,this resulted in a blurry image.This phenomenon has been identified by Wei and Levoy[5]Image analogies is a powerful tool in texture synthesis. Our application is just an extension of image analogies. Using scale invariant and orientation invariant key points, we may able to accelerate the algorithm by dividing the image to patches according to the key points,and then merge the patches,rather than synthesize the pixels one by one.Another possibility is to introduce the orientation invariant texture synthesis.A key point with a patch can be rotated in any direction.Therefore,it is capable to synthesize a new texture with different orientations. Finally,since SIFT can be used to match objects that appear with different positions,scales,and orientations in different images,it provides another possibility to extend the image analogies to handlefilters that warp the image. References[1]Alexei A.Efros and Thomas K.Leung.Texture syn-thesis by non-parametric sampling.In IEEE Interna-tional Conference on Computer Vision,Sep.1999. [2]Aaron Hertzmann,Charles E.Jacobs,Nuria Oliver,Brian Curless,and David H.Salesin.Image analo-gies.In Proceedings of SIGGRAPH,2001.[3]David G.Lowe.Distinctive image features fromscale-invariant keypoints.In Invernational Journal ofComputer Vision,2004.[4]Patrick Perez,Michael Gangnet,and Andrew Blake.Poisson image editting.In Proceedings of ACM SIG-GRAPH,2000.[5]Li-Yi Wei and Marc Levoy.Fast texture synthesis us-ing tree-structured vector quantization.In Proceed-ings of ACM SIGGRAPH,2003.4。