外文翻译中英文模板(新)2015.1

合集下载

外文文献翻译模板

外文文献翻译模板广东工业大学华立学院本科毕业设计（论文）外文参考文献译文及原文系部管理学部专业人力资源管理年级 2008级班级名称 08人力资源管理1班学号 150********学生姓名王凯琪指导教师2012年 5 月目录1 外文文献译文 (1)2 外文文献原文 (9)德国企业中老化的劳动力和人力资源管理的挑战本文的主要目的就是提供一个强加于德国公司的人力资源管理政策上的人口变化主要挑战的概况。

尽管更多方面的业务受到人口改变的影响，例如消费的改变或储蓄和投资，还有资金的花费，我们把注意力集中劳动力老龄化促使人事政策的变化上。

涉及广泛的人力资源管理政策，以有关进行创新和技术变化的招募问题为开端。

1 老化的劳动力及人力资源管理由于人口的变化，公司劳动力的平均年龄在未来将会更年长。

因此，劳动力高于50的年龄结构占主导地位的集团不再是一个例外，并将成为一个制度。

在此背景下，年长的工人的实际份额，以及最优份额，部分是由企业特征的差异加上外在因素决定的。

2 一般的挑战尽管增加公众对未来人口转型带来的各种挑战的意识，公司对于由一个老化劳动力引起的问题的意识仍然是相当低的。

事实上，只有25%的公司预计人口统计的变化在长远发展看来将会导致严重的问题。

然而，现在越来越多关于老化劳动力呈现的挑战和潜在的解决方案的文献。

布施提出了一种分析老员工一般能力的研究文集，并给出有关于年长工人的人力资源政策的实例。

目前，华希特和萨里提出一篇关于研究公司对于提前退休的态度和延长工作生涯的态度的论文。

在这些研究中，老员工的能力通常被认为是不同的，并不逊色，同时指出一个最优的劳动力取决于不同的公司的特殊要求。

一般来说，然而由于越来越缺少合格的员工，人口统计的变化将使得在各种人事政策方面上的压力逐渐增加。

特别是，没有内部人力资源部门的中小型企业，因此缺乏足够的特殊的基础设施，则面临着严峻的挑战。

与他们正常的大约两到五年的计划水平相反，他们将越来越多地要处理长期的个人问题和计划。

2015年职称英语新增文章翻译

其他部落比如霍皮和柯契地族也制作陶器。每个部落都用独特的方法制做出特别形式和样式的陶器。
亚瑟的创奇
很多文化都有某类能代表他们民族深信的价值观的英雄。有关亚瑟王的不同寻常的事情就是他的英雄主义传奇已经持续了几个世纪，并且影响力早已超越它所诞生的英国。在有关亚瑟王早期的故事中，他是一个战士，他在公元约700年抗击、震慑入侵的北欧人。许多这类亚瑟王的故事大概是基于事实。无论是否叫作亚瑟，大量证据证明是有这样一个战士存在的。后来这些故事被美化修饰，使得它们的真实性受到质疑。根据这些故事，我们知道亚瑟出生在位于英国西海岸廷塔杰尔的一座城堡里，这里经常狂风暴雨。由于亚瑟是国王UtherPendragon的私生子，所以他被巫师Merlin偷偷地带走，因此他不知道自己的真实身份。在王者之剑从它刺穿的石头中拔出时，他才成为国王。他娶了美丽的Guinevere，并且召集所有贵族骑士来到他的王宫，这其中就包括Lancelot，后来Guinevere背叛了亚瑟王和他，在一起。亚瑟王最终在决斗中被他的私生子Mordred打败，他的尸体被偷偷地运到阿瓦隆岛上。这个传奇故事对中世纪的英国人和法国人很有吸引力，那个时候骑士精神的道德标准——骑士的理想的品格——是许多故事中很重要的一部分。Galahad，Percival，Gawain，以及其他亚瑟王的骑士的英雄主义故事也都传播开来。在今天的英国，有许多地方都宣称是亚瑟王传奇遗址的一部分。廷塔杰尔还有一座成为废墟的城堡。格拉斯顿堡附近还有一座古代修道院的遗迹，据说亚瑟王和Guinevere的尸体在12世纪就在这里被挖掘出来。这些都不能证明传奇的真实性，但是它们却让这种神秘气氛延续下去。
AcrosstheDesert
穿越沙漠撒哈拉沙漠是世界上最大的沙漠。它从塞内加尔到埃及横跨非洲。撒哈拉沙漠的环境不好。白天非常热，晚上有时又很冷。在撒哈拉沙漠中很难找到水。 2006年，KevinLin,RayZahab和Charlie决定做些困难的事情。他们决定跑步穿越4300英里（6920千米）的撒哈拉沙漠。这似乎是不可能完成的，但是他们还想尝试一下。他们三人喜欢挑战自己，而这将是一个很大的挑战式 11月2日的早晨，Kevin,Ray和Charlie开始了他们跑步穿越撒哈拉的旅程。他们每天早晨5点开始跑，到上午11点停下来休息，然后到下午5点继续跑，一直跑到下午9点半。他们每天大概跑40英里（64千米）。每天如此，起床，跑步。听着iPod里的音乐不休停地跑。在旅途中，Kevin,Ray和Charlie需要吃很多事物。大多数的人每天需要2000卡路里的热量，而他们三人每天需要6000—9000卡路里。那真是很多食物！他们每天也需要喝大量的水。三人在途中也出现了很多问题，很多次他们都想放弃回家。白天通常很热（140华氏度／60摄氏度），高温导致他们生病，他们的腿和脚都受了伤。有时候天刮起了大风导致他们什么也看不见。有一次他们迷了路，但是他们没有放弃。111天以后，Kevin,Ray和Charlie成功完成了他们穿越撒哈拉沙漠的旅途。他们彼此拥抱，把手伸进红海的海水里，然后他们跑进旅馆好好洗了个澡。

5外文翻译模板(示例)

毕业设计(论文)外文资料翻译学院： (用三号楷体，下同）专业：姓名：学号：外文出处：附件：外文原文（复印件）(用外文写)译文标题（三号黑体，居中，单倍行距，段前0.5行；段后1行。

）ZUO Zheyi ,ZUO Zheer (四号Times New Roman ，行距20磅)(School of Electrical Engineering, Northeast Electric Power University, Jilin 132012, Jilin Province, China) （五号TimesNew Roman ，行距18磅，居中，段后0.5行，作者简介不用翻译成中文）摘要：(小四宋体，加粗) 文章摘要，要反映文章主要的技术内容，如研究目的、方法阐述和具体结论，阐明文章的行文脉络及文章的重点即可。

说明文章主要内容是什么，分哪几个部分介绍的，最后客观阐述结论或试验结果，不要有个人对研究结果的主观评价和研究背景介绍。

(小四号宋体，1.5倍行距，两端对齐)关键词：(小四宋体，加粗)油水两相流；经验模式分解；Elman 神经网络 (小四号宋体) 1 一级标题一（小四宋体，加粗，1.5倍行距，段前段后各0.5行） 1.1二级标题（小四宋体，加粗，1.5倍行距，段前段后各0.25行）正文×××××（小四号宋体，1.5倍行距）××…………。

（要求3000～5000汉字）正文中的公式。

221()mi i ij j E c t dt c ===∑⎰ (1)其中，ij c ( i =1,2,…,n ，j =1,2,…,m)表示IMF 的离散点的幅值，m 为信号采样点的数目。

（公式中涉及的字母所代表的物理量要全部指明。

公式所在行单倍行距，右对齐，符号大小在公式编辑器中尺寸栏中的“define ”下的“full ”项选12，其他默认）1.2 二级标题正文中的图。

外文翻译模板(1)

湖北知行学院金融专业英语外文文献译文本2014届原文出处:____ Companies Law Commons, SecuritiesLaw Commons_____________________ 译文题目：______调和税法和证券监管____院（系）经济与管理学院专业名称金融专业学生姓名王倩学生学号1211340073任课教师张真RECONCILING TAX LAW AND SECURITIES REGULATIONOmri Marian*Issuers in registered securities offerings must disclose the expected tax consequencesto investors investing in the offered securities (“nonfinancial tax disclosure”). ThisArticle advances three arguments regarding nonfinancial tax disclosures. First,nonfinancial tax disclosure practice, as the Securities and Exchange Commission(the SEC) has sanctioned it, does not fulfill its intended regulatory purposes. Cur-rently, nonfinancial tax disclosures provide irrelevant information, sometimes failto provide material information, create unnecessary transaction costs, and divertvaluable administrative resources to the enforcement of largely-meaningless require-ments. Second, the practical reason for this failure is the SECand tax practitioners’unsuccessful attempt to address investors’heterogeneous tax preferences. Specifi-cally, nonfinancial tax disclosure practice assumes the existence of a “reasonableinvestor”who is also an “average taxpayer,”and tax disclosures are drafted for thebenefit of this average taxpayer. The concept of an “average taxpayer,”however, isnot defensible. Third, the theoretical reason for the regulatory regime’s dysfunction-ality is the misapplication of mandatory disclosure theory to nonfinancial taxdisclosure requirements. Mandatory disclosure theory, even if accepted at facevalue, does not support the current regulatory framework, due to the special natureof tax laws. To remedy this failure, this Article describes the types of tax-relateddisclosures that mandatory disclosure theory would support. Under the proposedregulatory reform, nonfinancial tax disclosures will only includeissuer-level taxitems (namely, tax items imposed on the issuing entity) that affect how “reasonableinvestors”calculate their own individual tax liabilities. Under such a regime, thereis no need to rely on the “average taxpayer”construct. INTRODUCTION: APPLE’S BOND OFFERING AS AN ALLEGORYOn May 1, 2013, Apple, Inc. (Apple) made financial history wi its $17 billion bond offering,1 the largest-ever debt issuance by* Assistant Professor of Law, University of Florida Levin College of Law. For helpguidance, comments and critique, I am grateful to Jennifer Bird-Pollan, Stu Cohen, DaGamage, Joan Heminway, Michael Knoll, Leandra Lederman, Tom Lin, Randle PollaDexter Samida, Doug Shackelford, Danny Sokol, Emily Satterthwaite, and participantsconferences and workshops at the Northwestern University School of Law, the UniversityTennessee College of Law, the 2013 SEALS Annual Conference,the 2013 Law and SocAnnual Meeting, and the 8th Annual Junior Tax Scholars Workshop. For invaluable reseaassistance, I am indebted to Gus Gari. Any errors or omissions are my own.1. See generally Apple, Inc., Prospectus Supplement (Form 424B2), at S-15 (May 1, 20[hereinafter Apple’s Prospectus], available at /secfiling.cfm?filid=1193125-13-184506&cik=320193.12 University of Michigan Journal of Law Reform [VOL. 48:1 non-financial institution at the time.2 On page S-15 of the offeringdocument is a section titled “Certain U.S. Federal Income Tax Con-siderations.”3 This section provides information concerning the “U.S. federal income tax considerations of the ownership and dis-position”of the bonds.4Issuers in registered securities offerings are required to disclose all information that a reasonable investor would deem materialwhen making an informed investment decision.5 Since investorscare about their after-tax returns on investments,6 informationabout the tax costs associated with an investment could be consid-ered material.7 Indeed, registrants are required to disclose toinvestors all material tax consequences and to qualify the tax disclo-sure with an opinion.8 The tax opinion must address each materialtax issue discussed in the disclosure, express a legal conclusion about how the tax law applies to the facts of the particular offeringand its effect on investors’tax consequences, and explain the basis2. John Balassi & Josie Cox, Apple Wows Market With Record $17 Billion Bond Deal,REUTERS (Apr. 30, 2013), /article/2013/04/30/us-apple-debt-idUS BRE93T10B20130430. In September of the same year, VerizonCommunications smashedthis record with a $49 billion bond offering. John Atkins, Verizon Smashes Record with $49B, 8-Part Bond Offering, (Sept. 11, 2013), /sites/spleverage/2013/09/11/verizon-smashes-records-with-49b-8-part-bond-offe ring/ (last visited Aug. 21,2014).3. Apple Prospectus, supra note 1, at S-15.4. Id.5. See Zohar Goshen & Gideon Parchomovsky, The Essential Role of Securities Regulation,55 DUKE L. J. 711, 741 (2006). The disclosure documents the SEC regulates “disclose infor-mation about the companies’financial condition and business practices to help investorsmake informed investment decisions.”SEC, The Investor’s Advocate: How the SEC Protects Inves-tors, Maintains Market Integrity, and Facilitates Capital Information, /about/whatwedo.shtml (last visited Aug. 1, 2014); see also Goshen & Gideon, supra, at 740 (providingSEC disclosure regulations allow for greater “public disclosure”and “leads to fewer instancesof asymmetric information between traders”and more informed traders).6. See MYRON S. SCHOLES ET AL., TAXES AND BUSINESS STRATEGY 2 (3d ed. 2004) (discuss-ing why taxes influence investment decisions).7. See, e.g., SEC Staff Legal Bulletin No. 19, 2011 WL 4957889 11–13 (Oct. 14, 2011),available at /interps/legal/cfslb19.htm (discussing when tax conse-quences are “material”to investors) [hereinafter SEC Legal Bulletin]; see also William B.Barker, SEC Registration of Public Offerings Under the Securities Act of 1933, 52 BUS. LAW. 65,105–06 (1996) (discussing the proper disclosure of federal income tax consequences in regis-tered offering as part of a general discussion on the system of mandatory disclosure, which isintended to deliver investors with “accurate and current information”to support “fair andhonest securities market”). At the time of publication, Barkerwas a Senior Counsel to theSEC’s Division of Corporate Finance. William B. Barker, SEC Registration of Public OfferingsUnder the Securities Act of 1933, 52 BUS. LAW 65, 65 (1996).8. SEC Regulation S-K, 17 C.F.R. §229.601(b)(8) (requiring issuers to disclose to inves-tors the “material”tax consequences associated with purchasing, holding and disposing ofthe offered securitieFALL 2014] Reconciling Tax Law and Securities Regulation 3for such a conclusion.9 Apple’s tax disclosure section responds tothis regulatory framework.This Article suggests, however, that Apple’s tax disclosure in theoffering document does not provide any information that a “rea-sonable investor”would deem material. In fact, the disclosure provides little information at all, notwithstanding that the disclo- sure comprises four densely written pages. Specifically, the third sentence in Apple’s tax disclosure makes it clear that any tax conse-quences discussed therein are only applicable to investors purchasing the bonds in the initial offering.10 Investors in the sec-ondary market received no guidance concerning the tax consequences of investing in the bonds.In addition, Apple’s tax disclosure explicitly excludes certain classes of investors, including—among others—dealers in securi-ties, financial institutions, insurance companies, and other types ofinstitutional investors.11 It is well documented, however, that securi-ties in initial offerings are mostly allocated to the classes of institutional investors excepted from Apple’s tax disclosure. 12The result is rather remarkable: Apple’s tax disclosure does not describe the tax consequences to the investors that—as a practicalmatter—are expected to purchase the bonds in the initial offering.The tax disclosure also does not describe the tax consequences toany investor that purchases the bonds in the secondary market. Thelogical inference is that Apple’s tax disclosure section describes taxconsequences that are applicable to no one (or at least to only veryfew). It is hard to imagine, therefore, that Apple’s tax disclosure responds meaningfully to the rationales underlying mandatory tax9. SEC Legal Bulletin, supra note 7, at 12.10. See Apple Prospectus, supra note 1, at S-15 (“Except where noted, this summary dealsonly with a note held as a capital asset by a beneficial owner who purchases the note onoriginal issuance at the first price . . . .”).11. Id. (“This summary does not address all aspects of U.S. federal income taxes anddoes not deal with all tax consequences that may be relevant to holders in light of theirpersonal circumstances or particular situations, such as . . . tax consequences to dealers insecurities or currencies, financial institutions, regulatedinvestment companies, real estateinvestment trusts, tax-exempt entities, insurance companies and traders in securities thatelect to use a mark-to-market method of tax accounting for their securities.”).12. See SEC, Initial Public Offerings: Why Individual Investors Have Difficulty Getting Shares?,/answers/ipodiff.htm (last visited Aug. 1, 2014). There is ample evidencethat institutional investors are allocated most of the shares in IPOs (specifically on so called“hot”IPOs). See, e.g., Jay R. Ritter & Ivo Welch, A Review of IPO Activity, Pricing, and Allocations,57 J. FIN. 1795, 1808–15 (2002); see also Reena Aggarwal, Nagpurnanand R. Prabhala &Manju Puri, Institutional Allocations in Initial Public Offerings: Empirical Evidence, 57 J. FIN. 1421,1422 (2002) (finding that “institutions dominate IPO allocations”); Leland E. Crabbe &Christopher M. Turner, Does Liquidity of a Debt Issue Increase With Its Size? Evidence from theCorporate Bond and Medium-Term Note Markets, 50 J. FIN.1719, 1722 (1995) (finding both cor-porate bonds and medium-term notes “sold primarily to institutional investors”).s, and to support such disclosure with a legal opinion).调和税法和证券监管欧米- 玛丽安*证券发行的发行人在登记时必须披露预期的税收后果，提供给投资者投资于证券（“非金融税披露”）。

外文翻译及外文原文(参考格式)

外文翻译要求：1、外文资料与毕业设计（论文）选题密切相关，译文准确、质量好。

2、阅读2篇幅以上（10000字符左右）的外文资料，完成2篇不同文章的共2000汉字以上的英译汉翻译3、外文资料可以由指导教师提供，外文资料原则上应是外国作者。

严禁采用专业外语教材文章。

4、排序：“一篇中文译文、一篇外文原文、一篇中文译文、一篇外文原文”。

插图内文字及图名也译成中文。

5、标题与译文格式（字体、字号、行距、页边距等）与论文格式要求相同。

下页附：外文翻译与原文参考格式2英文翻译 (黑体、四号、顶格)外文原文出处：(译文前列出外文原文出处、作者、国籍，译文后附上外文原文)《ASHRAE Handbook —Refrigeration 》.CHAPTER3 .SYSTEM Practices for ammonia 3.1 System Selection 3.2 Equipment3.10 Reciprocating Compressors第3章氨制冷系统的实施3.1 系统选择在选择一个氨制冷系统设计时，须要考虑一些设计决策要素，包括是否采用（1）单级压缩（2）带经济器的压缩（3）多级压缩（4）直接蒸发（5）满液式（6）液体再循环（7）载冷剂。

单级压缩系统基本的单级压缩系统由蒸发器、压缩机、冷凝器、储液器（假如用的话）和制冷剂控制装置（膨胀阀、浮球阀等）。

1997 ASHRAE 手册——“原理篇”中的第一章讨论了压缩制冷循环。

图1.壳管式经济器的布置外文翻译的标题与译文中的字体、字号、行距、页边距等与论文格式相同。

英文原文(黑体、四号、顶格)英文翻译2（黑体，四号，顶格）外文原文出处：（黑体，四号，顶格）P. Fanning. Nonlinear Models of Reinforced and Post-tensioned Concrete Beams. Lecturer, Department of Civil Engineering, University College Dublin. Received 16 Jul 2001.非线形模型钢筋和后张法预应力混凝土梁摘要:商业有限元软件一般包括混凝土在荷载做用下非线性反应的专用数值模型。

高考英语满分作文及翻译修订稿

高考英语满分作文及翻译Document number【AA80KGB-AA98YT-AAT8CB-2A6UT-A18GG】2015高考英语满分作文范文及翻译（1）假设你叫王明，昨天收到了笔友David的e-mail，得知他不久要到北京来学习中文。

他想了解如何学好中文。

请你用英文给他回复一封e-mail，介绍学习中文的体会和方法，提出你的建议，以及表达你帮助他学好中文的愿望。

【例文】Dear David,I'm glad you'll come to Beijing to learn Chinese. Chinese is very useful, and many foreigners are learning it now. It's difficult for you because it's quite different from English. You have to remember as many Chinese words as possible. It's also important to do some reading and writing. You can watch TV and listen to the radio to practise your listening. Do your best to talk with people in Chinese. You can learn Chinese not only from books but also from people around you. If you have any questions, please ask me. I'm sure you'll learn Chinese well.Hope to see you soon in Beijing.Yours,Wang Ming【翻译】亲爱的大卫，我很高兴你会来北京学习中文。

(论文)外文翻译模版

沈阳工业大学
本科生外文翻译
文章中文题目：_______________________________ 文章外文题目：_______________________________
学院：
专业班级：
学生姓名：
指导教师：
年月日
外文翻译格式、装订要求（对中文的要求）：
1．字数要求：不少于3000中文字符，或不少于10000个英文字符。

2．外文翻译内容一律用A4纸打印；
3．图表5号字，宋体。

图表要居中；图的编号和名称在图下，居中；表的编号和名称在表上，居中。

4．大标题用三号宋体（加粗）、小标题用四号宋体（加粗）、内容用小四宋体、行间距20磅、页边距上下2.54厘米、左右3.17厘米。

5.中文在上，外文在下，左侧装订。

（外文必须用A4纸打印，原文其它格式格式可以不变）
6.统一用上页封面和指导教师评语，指导教师评语装订在最后一页。

1-2015届本科毕业论文外文翻译模板(说明版)

1-2015届本科毕业论文外文翻译模板(说明版)
本科生毕业设计 (论文)
外文翻译
原文标题 Our Country Commercial Bank Financial
Control’s Reform Innovates 译文标题我国商业银行的财务管理的改革创新
作者所在系别会计系作者所在专业财务管理作者所在班级 B11622 作者姓名张昊作者学号 20114062235 指导教师姓名李娅捷指导教师职称教授完成时间
2014
年
12
月
30日
北华航天工业学院教务处制
整篇文档通用的页面设置参数为：A4打印纸（29.7厘米×21厘米）。

英文及阿拉伯数
本页空白单元格格式均为：宋体（
汉
字），小
注：1. 指导教师对译文进行评阅时应注意以下几个方面：①翻译的外文文献与毕业设计（论文）的主题是否高度相关，并作为外文参考文献列入毕业设计（论文）的参考文献；②翻译的
外文文献字数是否达到规定数量（3 000字以上）；③译文语言是否准确、通顺、具有参考价值。

2. 外文原文应以附件的方式置于译文之后。

中外翻译模板英语作文

中外翻译模板英语作文英文回答：Template for Translation between Chinese and English。

Introduction。

Translation is a complex and challenging task that requires a deep understanding of both the source and target languages. In order to produce accurate and fluent translations, it is essential to use a structured approach that takes into account the linguistic and cultural differences between the two languages. This template provides a step-by-step guide to translating between Chinese and English, with a focus on preserving the meaning and tone of the original text.Step 1: Understand the Source Text。

Begin by carefully reading the source text andidentifying the main ideas and supporting points. Pay attention to the tone and style of the writing, as well as any cultural references or idioms that may require special attention. Use a dictionary or online resources to look up any unfamiliar words or phrases.Step 2: Analyze the Target Language。

优秀翻译模板作文英语版

优秀翻译模板作文英语版Excellent Translation Template Essay in English。

Translation is an art that requires both linguistic proficiency and cultural understanding. It is the bridge that connects people from different countries and allows them to communicate and understand each other's thoughts and ideas. In today's globalized world, the demand for professional translators is higher than ever, as businesses, governments, and individuals seek to overcome language barriers and reach new markets. In this essay, we will explore the qualities of an excellent translation, the challenges faced by translators, and the importance of translation in the modern world.An excellent translation is more than just converting words from one language to another. It requires a deep understanding of the source language and the target language, as well as the ability to convey the original meaning, style, and tone of the text. A good translator must have a strong command of both languages, as well as a keen eye for detail and an understanding of the cultural nuances that may affect the meaning of the text. They must also be able to adapt the translation to the target audience, whether it is a legal document, a marketing brochure, or a literary work. A successful translation is one that reads as if it were originally written in the target language, without losing the essence of the original text.One of the biggest challenges faced by translators is the ambiguity and complexity of language. Words and phrases often have multiple meanings and interpretations, and it is the translator's job to choose the most appropriate one in the given context. This requires not only linguistic knowledge but also critical thinking and problem-solving skills. Additionally, translators must be aware of the cultural, historical, and social context of the text, as well as the intended audience, in order to accurately convey the message of the original text. Furthermore, translators must keep up with the evolving nature of language, as new words and expressions constantly emerge, especially in the age of technology and social media.The importance of translation in the modern world cannot be overstated. In the business world, accurate and culturally sensitive translations are essential for companies to expand into new markets and reach a global audience. In the legal field, translations are crucial for international agreements, contracts, and court proceedings. In the medical and scientific fields, accurate translations of research papers and medical documents are essential for the global exchange of knowledge and advancements. In literature and the arts, translations allow people to access and appreciate the works of writers and artists from around the world. In diplomacy and international relations, translations are vital for effective communication between countries and cultures.In conclusion, an excellent translation requires a combination of linguistic expertise, cultural understanding, and adaptability to the target audience. Translators face numerous challenges, from the ambiguity and complexity of language to the ever-changing nature of words and expressions. However, the importance of translation in the modern world cannot be overstated, as it is essential for global communication, business expansion, knowledge exchange, and cultural appreciation. As the world becomes increasingly interconnected, the demand for professional translators will continue to grow, and their role in bridging language barriers will become even more crucial.。

2015版英文翻译模板

英文翻译姓名：张衡学号： 1102080234指导教师：孙中桥专业：信息管理与信息系统班级： 2011级时间：2015年6月25日信息管理系信息管理系英文翻译评价表学生姓名张衡性别男学号1102080234外文文献标题Fast Component-Based QR Code Detection in Arbitrarily AcquiredImages外文文献出处以下内容由指导教师填写（打勾“√”选择）评价项目评价结论打勾评价结论打勾评价结论打勾是否外文期刊文献是否与本人论文相关完全相关一般不相关翻译工作量超负荷饱和不饱和翻译态度认真一般不认真翻译进度按计划执行一般未按计划执行翻译训练效果优良中差综合评语（是否完成了规定任务、效果是否符合要求等）指导教师签名：2015年4 月25日制表：李铁治注1：此表与翻译文本一起装订；注2：为了加强学生外语应用能力的训练，每位同学至少选择毕业论文中一篇外文参考文献（10000英文字符），翻译成中文。

外文文献及中文译文不装钉进论文中，只形成单行本放入档案袋即可。

英文原文Fast Component-Based QR Code Detection in Arbitrarily Acquired ImagesAbstract Quick Response (QR) codes are a type of 2D bar- code that is becoming very popular, with several application possibilities. Since they can encode alphanumeric charac- ters, a rich set of information can be made available through encoded URL addresses. In particular, QR codes could be used to aid visually impaired and blind people to access web based voice information systems and services, and au- tonomous robots to acquire context-relevant information. However, in order to be decoded, QR codes need to be prop- erly framed, something that robots, visually impaired and blind people will not be able to do easily without guid- ance. Therefore, any application that aims assisting robots or visually impaired people must have the capability to de- tect QR codes and guide them to properly frame the code. A fast component-based two-stage approach for detecting QR codes in arbitrarily acquired images is proposed in this work. In the first stage, regular components present at three corners of the code are detected, and in the second stage ge- ometrical restrictions among detected components are veri- fied to confirm the presence of a code. Experimental results show a high detection rate, superior to 90 %, at a fast speed compatible with real-time applications.Keywords QR code • Component-based detection • Haar-like features • Cascade classifier 1 IntroductionCompared to traditional 1D barcodes, 2D barcodes can en- code a larger amount of data, including alphanumeric char- acters. QR code, which stands for Quick Response Code, is a type of two-dimensional code introduced by Denso Wave in 1994 [12]. They have been designed to be easily found and to have its size and orientation determined under bad imaging conditions. In addition, ISO/IEC 18004 specifies an error correction scheme that can recover at most 30 % of occluded or damaged symbol area. These features make the QR code an extremely well succeeded technology in the area of barcodes. Figure 1 shows some examples of QR codes.Fig. 1 Samples of QR codeQR codes were initially used by the automotive industry to track vehicle parts during the manufacturing process [12]. Nowadays, QR codes are most commonly used as ―physical hyperlinks‖ (encoded hyperlinks), notably in the advertising industry, to connect places and objects to websites contain- ing additional context relevant information. Applications in education and entertainment [8, 23, 25], data and system se-curity [9, 16, 19], specific service offer [7, 27], among others are also emerging. Another possible application consists in helping visually impaired and blind people, or even robots, in several tasks such as indoor navigation, shopping, read- ing, and much more [2, 11, 13].The existing decoders, easily found for mobile devices, are able to work correctly only if codes are properly framed, with code region corresponding to at least 30 % of the im- age. When exploring an environment, visually impaired or robots will not be able to capture such images unless they are told where those codes are located. Thus, in order to make useful applications for them viable, detecting the presence of a code in an image is a necessary step prior to the decod- ing process.In related literature, the majority of work that mention QR code recognition or detection is actually concerned with improving image quality or determining the exact contours of the code rather than finding (i.e, deciding whether there is or there is not) a QR code symbol in an image [10, 20, 22]. In fact, most of the images considered in those works areimages acquired with the specific intent of capturing only the code symbol.Some few works that deal with the problem of finding QR codes propose solutions that rely on auxiliary informa- tion such as visual cues or RFID tags [11, 28]. Although such approach is possible in controlled environments, it is not practical in general contexts.This work addresses the problem of detecting QR codes in arbitrarily acquired images without relying on auxiliary cues. The aim is not only to detect the presence of code symbols, but also to delimit their position within an image as accurately as possible. That would allow, additionally, to instruct an user or robot to properly approach the camera towards the code.To that end, a component-based, two-stage detection ap- proach is proposed. In the first stage, a square pattern, called finder pattern (FIP), located at three corners of any QR code is detected using a cascaded classifier trained according to the rapid object detection method proposed in [26]. In the second stage, geometrical relationships among detected can- didate finder patterns are analyzed in order to verify if there are subgroups of three of them spatially arranged as three corners of a square region. A report on preliminary results of this approach has been previously published in [3]. In this paper, a reformulated report of the approach and an extended set of results are presented. In particular, a detailed account of the second stage, including a formal description of the component aggregation algorithm, new results obtained with a larger training and test sets, discussions concerning addi- tional parameters that have not been examined in the previ- ous work, and quantitative results, not presented before, on QR code detection performance are presented.This text is organized as follows. In Sect. 2, important structural aspects of QR codes which will be explored in the proposed approach are first described. Then, an overview of the proposed two-stage approach is presented, together with a brief review on Viola-Jones method for rapid object de- tection, used in the first stage of the approach. In Sect. 3, the problem of aggregating information obtained in the first stage of the detection process is expressed in graph-based formalism and a simple algorithm to solve the problem is proposed. The procedure to determine appropriate values for the training parameters of the FIP detector, in an OpenCV implementation of Viola-Jones’ method, is discussed and described in Sect. 4. Then, in Sect. 5 experimental results to assess influence of some parameters in FIP as well as QR code detection, and frame rates achieved in video process- ing are reported. Although extensive experimental variation was carried out, not all possibilities arising from parameter variations were evaluated. Nevertheless, QR code detection rate superior to 90 % was observed on test images, suggest- ing that even better rates can be achieved. Finally, in Sect.6 a summary of the main contributions of this work and some issues for future investigation are listed.2 Component-Based Approach for QR Code DetectionIn this section, some structural aspects of QR codes and detection requirements in a real-time application scenario that have motivated the proposed two-stage approach are first described. Then, an overview of the proposed approach followed by a brief review on Viola-Jones’ method for object detection is presented.2.1 QR CodeA QR code symbol, according to the ISO/IEC standard 18004, has a general structure that comprises data, version information, error correction code words, and the following regions illustrated in Fig. 2:Fig. 2 QR code structure (FIP: Finder Pattern / TP: Timing Pattern /AP: Alignment Pattern)– a quiet zone around the symbol,–three finder patterns (FIP) in the corners,–two timing patterns (TP) between the finder patterns, and– a certain number of alignment patterns (AP) inside the data areaThe internal organization of a code with respect to its structural components may vary depending on the amount of data encoded. The smallest units (black or white squares) that compose a QR code are called modules. There are various versions of the symbols, from 1 to 40, having distinct information storage capabilities that are determined by predefined number of modules in symbol area. Figure 3 shows an example of a version 1 code, highlighting the number of modules. Figure 4 shows additional examples that illustrate the relation between number of modules and number of encoded characters. Note that the alignment patterns occur more times in larger versions of the code.Fig. 3 QR code version 1 is composed of 21×21 modulesFig. 4 Different versions of QR code symbols and respective number of encoded alphanumeric charactersThere is a visually distinctive pattern at three corners of QR codes, known as finder pattern (FIP). FIPs have been specially designed to be found in any search direction as the sequence of black (b) and white (w) pixels along any scan line that passes through its center preserve the special sequence and size ratio w:b:w:bbb:w:b:w, as can be seen in Fig.5.Fig. 5 Black and white pixel proportions in finder patterns for diagonal (d), vertical (v) and horizontal (h) scan lines2.2 Detection RequirementsOne of the first requirements considering real-time applications is fast detection. Moreover, since there is no prior knowledge about the size of a code, another important feature of the detector is scale invariance. In addition, in order to succeed in uncontrolled environments, the detector should be also robust with respect to variation in illumination, rotation, and perspective distortion.In order to establish some limits to the detection task, in this work attention is paid to scale invariance and robustness with respect to variation in illumination condition, while rotations and perspective distortions are assumed to be small.Further than that, it is assumed that QR codes are printed in black and white (or at least dark foreground and light background). Color is not an issue since all processing considers gray tone images.2.3 Overview of the Proposed ApproachDirect detection of the whole code may not be an easy task because the internal structural organization may vary from code to code, depending on code version. Moreover, extraction of features of the code area, such as histograms or frequency information, can be greatly affected by image resolution, code scale, or noise. For instance, algorithms for recognition of QR codes frequently rely on the regularity of scan lines over the FIPs (see Fig. 5) to find them and then determine the exact contours of the code. However, that regularity may be disrupted in low quality images. Thus, in order to deal with unconstrained images, the processing algorithm should not depend on fine details.On the other hand, FIPs are typically the largest structures within a code, have a fixed shape, and appear in exactly three corners of all codes. These characteristics of FIPs suggest a component-based detection approach to the problem of detecting QR codes. The main idea of component-based detection [21] is the detection of parts of the object and then the analysis of detected parts in order to find a coherent arrangement forming the integral object. Arrangement of the parts can take into consideration geometrical restrictions. When the relation between parts can not be easily stated, machine learning based approaches can be used as in [1, 14].Taking the above observations into consideration, the proposed approach for detecting QR codes consists of a twostage method. In the first stage the aim is to detect FIPs, and in the second stage the goal is to integrate information about detected FIPs and decide whether subsets of three of them correspond to a QR code or not.In the first stage, it is desirable to maximize true positives (TP) while maintaining a controlled number of false positives (FP). Thus, for the detection of FIPs, some mechanism to control TP and FP should be available, and the detection process should fulfill the requirements listed above. In particular, it should be fast and invariant to scale and variation of illumination conditions. To meet these requirements, Viola-Jones’ rapid object detection method [26] is proposed for the first stage.In the second stage, the algorithm should be able to find sets of three FIPs that are the corners of a same QR code. To that end, geometrical restrictions as well as size related restrictions are considered, as detailed in Sect. 3.2.4 A Review on Viola-Jones’ Object Detection FrameworkViola-Jones’ object detection framework has the following characteristics:–it uses simple features (Haar-like);–it performs fast computation of features using integral image;–it is based on a cascaded approach to train a classifier: each stage is trained in such a way as to achieve fixed TP and false alarm (FA) rates, using a boosting algorithm;–it is very fast in practice since features are computed rapidly and the cascade is designed to discard the majority of negative samples in its early stages, eliminating the need for calculating responses of all stages to every sample considered.The Viola-Jones framework can be illustrated with the following concise description: The detection process consists in acquiring a complete sample set from the input image and submitting each sample to a cascade classifier whose stages have been trained by a boosting scheme that aggregates weak classifiers made of Haar-like features. These features are calculated in constant time using a matrix called integral image. Below some key concepts are reviewed and the detection process is explained.Sample set A sample consists of a sub-region of the image restricted to a window and the complete sample set of the image is obtained by sliding the window on the image and varying its size from a given minimum to the largest possible size in the image.Cascade classifier A cascade consists of a series of consecutive classifiers (stages) trained to reject samples that do not match the searched pattern while accepting the majority of positive samples and passing them to the next stage. A sample is said to be detected when it is accepted from the first to the last cascade stages, without rejection.Each stage is built from a set of weak classifiers aggregated in a committee by a boosting algorithm and can be seen as an independent classifier designed to obtain a very high hit rate (typically 99 % or more) with an acceptable false alarm rate (typically between 40 % and 60 %). Figure 6 illustrates the concept of a cascade classifier.Fig. 6 Illustration of the cascade approach to detectionBoosting Boosting works in rounds, iteratively training a set of weak classifiers while reweighting the training samples so that ―hard‖ samples will have increased relative importance in the set. In each round the best weak classifier is selected to compose the resulting classifier [15]. Thus, boosting can achieve the specified hit rate and false alarm rate as it increases the weak classifiers in the combination (as long as the features used have enough representational power). Every weak classifier is typically the result of the computation of a Haar-like feature followed by a binary threshold, although they can have more sophisticated forms like simple trees.Haar-like features The features used by the classifiers proposed in [26], inspired from [24], are based on feature prototypes shown in Fig. 7. In [18], this set has been extended with additional 45 degree rotated prototypes, shown in Fig. 8.Fig. 7 Basic feature prototype setFig. 8 Additional rotated feature prototypes in the extended setGiven a square window, all features used are derived from these prototypes by defining its width, height and relative position within the window. Some examples are shown in Fig. 9. Note, for instance, that features in Fig. 9(a) and Fig. 9(b) are both based on the same prototype.Fig. 9 Examples of Haar-like featuresFeature values are computed with respect to a sample that corresponds to a sub-region of the image under a sliding evaluation window. By definition, the value of a feature for each sample is the sum of the pixel values in the white rectangle area subtracted from the corresponding summation in the black rectangle area.3 Aggregation algorithmOnce FIP candidates are detected in the first stage, arrangement of three candidate FIPs that are likely to correspond to the corners of a QR code must be examined in the second stage. In this section, an algorithm that considers size, distance and angle restrictions to perform such examination is proposed. Note that besides these geometrical restrictions, additional information such as the number of overlapping detections or texture around the FIPs could be also used to rule out some subsets or support others. However, taking into consideration the requirement for fast processing, only size, distance, and angle information are considered.The proposed solution for aggregating FIP candidates is presented in Algorithm 1. Lines 1 to 3 correspond to vertex set creation. Lines 4 to 13 correspond to the creation of edges, whenever a pair of vertices satisfy the size and the distance criteria, with respective tolerances. The loop structure from line 14 to 23 is the main part of the algorithm, where cliques of size three whose edges pairwise satisfy the orientation criterion are found. For each edge (u, v), all edges (u, v) that are adjacent to one of its extremities (u) are verified. For each of such pairs of edges (i.e., {(u, v), (u, v)}), first the existence of the third edge (i.e., edge (v, v)) is verified. If the third edge is not there, that means that either the two vertices,v and v do not satisfy the size or distance criteria, or that the third edge has already been processed (in this case, the triplet {u, v, v} is already in the output list). If the third edge (v, v) is present, and if the first two satisfy the orientation condition, then the subset of the three vertices {u, v, v} is added to the output list. Note that there is no need to examine edges that are adjacent to the other extremity (v) since those will be evaluated later. The algorithm endswhen all edges are processed.4 Procedure for Training and Evaluating FIP DetectorsClassifier training and evaluation tools that implement the Viola-Jones’ method are available respectively as opencv-haartraining and opencv-performance utilities in OpenCV 2.0 [4]. For the training of a cascaded classifier, a set of positive samples (cropped image regions containing only a target object) and negative background images (with no target objects) should be provided. For each stage, the training algorithm randomly selects negative samples from the background images that have been misclassified (as positives) by the preceding stages.Another utility available in OpenCV is opencv-createsamples, that allows application of transformations on positive samples with controlled random variations in intensity, rotation and perspective distortion. Examples of transformed samples obtained from a positive sample are shown in Fig. 13.Fig. 13 Examples of samples generated by random transformations on a positive sample4.1 Main Training ParametersThe training process implemented in opencv-haartraining involves several parameters that need to be tuned for each application. Among them, the following parameters have been considered in this work:–Feature set: It can be the basic set shown in Fig. 7 (MODE = Basic) or the basic set in conjunction with the extended set shown in Fig. 8 (MODE = Extended/All).–Symmetry: When the target pattern is symmetrical, the training algorithm can be restricted to consider only half part of the positive samples (SYM = Symmetric/Y) in order to reduce processing time during training, or not (SYM= Asymmetric/N).–Classifier topology: Rather than cascaded stages (MTS = Cascade), it is possible to allow splits (MTS = Tree) that turns the classifier into a simple CART tree [5].–Weak classifier number of splits: A weak classifier in its simplest form is just a single Haar-like feature and a binary threshold. They are combined with other weak classifiers by the boosting algorithm to form a strong classifier corresponding to one stage of the cascade. It is possible to allow weak classifiers to learn more complex relationships in the training pattern by letting them to be not just a single feature (NS = 1) but a simple CART tree with a limited small number of splits (NS > 1). For face detection, empirical observations [17] indicate that there is an increase in cascade performance when splits in the weak classifiers are allowed.–Maximum false alarm rate / Minimum hit rate: Each cascade stage must comply with a maximum false alarm rate (FA) and with a minimum hit rate (HR) for the samples supplied to it in order to be considered trained. A too low false alarm rate requirement may cause the stage to become overly complex diminishing the benefits of the cascade approach. A very high minimum hit rate restriction may have a similar effect.–Number of training samples: The number of positive training samples (SAMPLES).–Size of training samples: All training samples are resized to a fixed dimension (SIZE). For face detection, it has been observed that 20×20 is an appropriate size [17].4.2 Detection Evaluation MetricsA classifier trained using the opencv-haartraining utility can be evaluated using the opencv-performance utility. The performance evaluation utility requires test images with ground-truth data (i.e., positions of the target samples). Such images can be artificially generated by the opencv-createsamples utility, that inserts one target sample in each background image. This, in conjunction to opencv-performance, is useful for automating the performance tests, by providing to them a test set of positive samples and negative background images as the ones provided for training. Figure 14 shows an example of target sample inserted into a background image.Fig. 14 A background image with a FIP inserted around its central positionScaling factor For each position of the image, several samples are obtained by varying the window size from a minimum size to the maximum possible size in that position. Since all samples are resized to the size of the detector, this minimum size should be the SIZE of the detector (i.e. the size of the samples used for training the detector). The scaling factor (SF) determines how window size varies for sampling. For instance, SF = 1.10 means that the size of the processing window will be increased in 10 % at each sampling round in a same position. The larger the scale factor, less samples are considered in each position. Note that the classifier can be considered scale invariant due to this resizing of all samples to a fixed size.Minimum overlapping detections Often more than one sample among those corresponding to small shifts of the window around the target object result in positive detection. In such cases, it is convenient to establish some criteria in order to consolidatemultiple redundant detections into a single detection representing all of them. Naturally, the larger the number of detections that overlap in a given neighborhood, the larger the evidence that there is a target object inthat position. Therefore, a minimum number of overlaps (ND) in the set of overlapping detections can be established as determining a single detection, that is, a set of overlapping detections is considered a detection if and only if the number of detections in the set is at least ND.In the implementation used in this work, two detections are considered overlapping if their position in x and in y coordinates differ by no more than 20 % of the width of the smaller detected sample and their widths differ by no more than 20 %.4.3 Estimation of Training Parameter ValuesViola-Jones framework is widely known and very popular for face detection and the literature concerning the training of the cascade classifier often recommends training practices, parameters and sets, suitable for that application domain [6]. Since FIPs are simpler and much different from face patterns, variation of parameter values for the cascade training of FIP detectors is a subject of study in this work.Due to the large number of parameters in the training process, an exhaustive assessment of possible combinations of parameter values is not feasible. The approach for determining the parameter values used in this work for the cascade training of a FIP detector is similar to the one used in [17] for face detection.The methodology consists in first establishing a sequential order as well as initial values for the parameters to be evaluated. Then, following the established order, the effect of individual parameter variation is assessed while the values of all other parameters are kept fixed. After assessment of a parameter, its initial value is replaced by the best value found, and the assessment proceeds to the next parameter. Although this analysis cannot be considered exhaustive, it provides a very good indication of the parameter variation influence in final detection performance while avoiding the combinatorial explosion that would arise from experimenting all possibilities.Parameter values for training FIP detectors were chosen following the strategy described above. A base set consisting of 380 FIP samples cropped from a variety of different images, and a set of 1500 background images (without QR codes) for the negative samples were divided into training and test sets. All positive samples for training were generated from the 285 FIPs in the base set, applying transformations using the opencv-createsamples utility with intensity variation limited to ±80 and rotation limited in x,y to ±23◦ and in z (perspective) to ±34◦. The test samples were generated from the remaining 95 FIPs applying the same variation in intensity and in x and y, and in z limited to ±23◦. The background images were split into two subsets of equal size (750 images each), being used respectively for training and testing.Initial values of parameters were set to HR = 0.998, ST = 15, SYM = N (Asymmetric), MTS = Cascade, NS =2, FA = 0.5, SAMPLES = 4000 (positives and negatives,4000 each), and SIZE = 20×20. Recall and FP shown for each training is relative to the test set, computed as described in Sect. 4.2 using detector application parameters SF = 1.10 and ND = 1. Dark nodes indicate the values chosen for the respective parameters. Beyond considering just recall and FP, the choice of the best parameter values were made taking into account the simplicity of the resulting classifier and the balance of its cascade (i.e. uniformity of feature quantity among consecutive stages). Moreover, recall values were prioritized over FP because while the FIP aggregation algorithm can not recover undetected FIPs, it does can disregard false FIP detections.5 Experimental ResultsIn this section, experimental results regarding FIP and QR code detection in still images, and frame rates achieved for QR code detection in video images are reported. Some images that illustrate detection results are presented. Additional resulting images can be accessed from URL p.br/demos.5.1 FIP DetectionAfter choosing training parameter values as described in Sect. 4.3, a FIP detector was trained with the chosen parameter values, using a set of 462 positive samples (cropped FIP regions) and 820 background images with no QR codes from where 4000 negative exemplars were randomly sampled for training each stage.The trained FIP detector was evaluated on a validation set A consisting of 74 images containing 222 positive samples (exactly one complete QR code in each image), totally distinct from those in the training set. During this evaluation, the influence on detection performance of two detector application parameters (see Sect. 4.2), scaling factor (SF) and number of overlapping detections (ND), have been analyzed as described below.Scaling factor Figure 17 shows that the scaling factor (SF) directly impacts the quality of results. A wide range of result quality was observed from SF = 1.05 (recall: 0.98 / FP:628) to SF = 1.30 (recall: 0.86 / FP: 93). As the scaling factor is incremented, both recall and FP decrease. Considering that high recall and small FP are desirable, scaling factor of 1.10 as highlighted in the chart (recall: 0.99 / FP: 290) was adopted for the successive experiments.Fig. 17 Recall and FP of the trained FIP detector varying the scaling factor used during FIP detection。

外文翻译原文模板

1、外文资料翻译内容要求：外文资料的内容应为本学科研究领域，并与毕业设计（论文）选题相关的技术资料或专业文献，译文字数应不少于3000汉字以上，同时应在译文末注明原文的出处。

不可采用网络中直接有外文和原文的。

2、外文资料翻译格式要求：译文题目采用小二号黑体，居中；译文正文采用宋体小四号，段前、段后距为0行；行距：固定值20磅。

英文原文如果为打印的话用新罗马（Times New Roman）小四号字。

装订时原文在前，译文在后。

文章中有引用的地方在原文中也要体现。

参考文献也要翻译成中文！An Energy-Efficient Cooperative Algorithm for Data Estimation inWireless Sensor NetworksAbstract – In Wireless Sensor Networks (WSN), nodes operate on batteries and network’s lifetime depends on energy consumption of the nodes. Consider the class of sensor networks where all nodes sense a single phenomenon at different locations and send messages to a Fusion Center (FC) in order to estimate the actual information. In classical systems all data processing tasks are done in the FC and there is no processing or compression before transmission. In the proposed algorithm, network is divided into clusters and data processing is done in two parts. The first part is performed in each cluster at the sensor nodes after local data sharing and the second part will be done at the Fusion Center after receiving all messages from clusters. Local data sharing results in more efficient data transmission in terms of number of bits. We also take advantage of having the same copy of data at all nodes of each cluster and suggest a virtual Multiple-Input Multiple-Output (V-MIMO) architecture for data transmission from clusters to the FC. A Virtual-MIMO network is a set of distributed nodes each having one antenna. By sharing their data among themselves, these nodes turn into a classical MIMO system. In the previously proposed cooperative/virtual MIMO architectures there has not been any data processing or compression in the conference phase. We modify the existing VMIMO algorithms to suit the specific class of sensor networks that is of our concern. We use orthogonal Space-Time Block Codes (STBC) for MIMO part and by simulation show that this algorithm saves considerable energy compared to classical systems.I. INTRODUCTIONA typical Wireless Sensor Network consists of a set of small, low-cost and energy-limited sensor nodes which are deployed in a field in order to observe a phenomenon and transmit it to a Fusion Center (FC). These sensors are deployed close to one another and their readings of the environment are highly correlated. Their objective is to report a descriptive behavior of the environment based on all measurements to the Fusion Center. This diversity in measurement lets the system become more reliable and robust against failure. In general, each node is equipped with a sensing device, a processor and a communication module (which can be either a transmitter or transmitter/receiver).Sensor nodes are equipped with batteries and are supposed to work for a long period of time without battery replacement. Thus, they are limited in energy and one of the most important issues in designing sensor networks will be the energy consumption of the sensor nodes. To deal with this problem, we might either reduce the number of bits to be transmitted by source compression or reduce the required power for transmission by applying advanced transmission techniques while satisfying certain performance requirement.A lot of research has been done in order to take advantage of the correlation among sensors’ data for reducing the number of bits to be transmitted. Some are based on distributed source coding[1]while others use decentralized estimation[2-5]. In [1], authors present an efficient algorithm that applies distributed compression based on Slepian – Wolf[14] encoding technique and use an adaptive signal processing algorithm to track correlation among sensors data. In [2-5] the problem of decentralized estimation in sensor networks has been studied under different constraints. In these algorithms, sensors perform a local quantization on their data considering that their observations are correlated with that of other sensors. They produce a binary message and send it to the FC. FC combines these messages based on the quantization rules used at the sensor nodes and estimates the unknown parameter. Optimal local quantization and final fusion rules are investigated in these works. The distribution of data assumed for sensor observation in these papers has Uniform probability distribution function. In our model we consider Gaussian distribution introduced in [17] for sensor measurements which ismore likely to reality.As an alternative approach, some works have been done using energy-efficient communication techniques such as cooperative/virtual Multiple-Input Multiple-Output (MIMO) transmission in sensor networks [6-11]. In these works, as each sensor is equipped with one antenna, nodes are able to form a virtual MIMO system by performing cooperation with others. In [6] the application of MIMO techniques in sensor networks based on Alamouti[15] space-time block codes was introduced. In [8,9] energy-efficiency of MIMO techniques has been explored analytically and in [7] a combination of distributed signal processing algorithm presented and in [1] cooperative MIMO was studied.In this paper, we consider both techniques of compression and cooperative transmission at the same time. We reduce energy consumption in two ways; 1) processing data in part at the transmitting side, which results in removing redundant information thus having fewer bits to be transmitted and 2) reducing required transmission energy by applying diversity and Space-Time coding. Both of these goals will be achieved by our proposed two-phase algorithm. In our model, the objective is to estimate the unknown parameter which is basically the average of all nodes’ measurements. That is, exact measurements of individual nodes are not important and it is not necessary to spend a lot of energy and bandwidth to transmit all measured data with high precision to the FC. We can move some part of data processing to the sensors side. This can be done by local data sharing among sensors. We divide the network into clusters of ‘m’ members. The number of members in the cluster (m) is both the compression factor in data processing and also the diversity order in virtual-MIMO architecture. The remaining of this paper is organized as following: in section II we introduce our system model and basic assumptions. In section III we propose our collaborative algorithm. In section IV we present the mathematical analysis of the proposed algorithm and in section V we give some numerical simulations. Finally section VI concludes the paper.II. SYSTEM MODELA. Network ModelThe network model that we use is similar to the one presented in [2-5].Our network consists of N distributed Sensor Nodes (SN) and a Fusion Center (FC). Sensors are deployed uniformly in the field, close to one another and each taking observations on an unknown parameter (θ). Fusion Center is located far from the nodes. All nodes observe same phenomenon but with different measurements. These nodes together with the Fusion Center are supposed to find the value of the unknown parameter. Nodes send binary messages to Fusion Center. FC will process the received messages and estimate the unknown value.B. Data ModelIn our formulation we use the data model introduced in[17]. We assume that all sensors observe the same phenomenon (θ) which has Gaussian distribution with variance σx 2. They observe different versions of θ and we model this difference as an additive zero mean Gaussian noisewith variance σn 2. Therefore, sensor observations will be described byn i i θx += （1） Where θ ~ N (0, σx 2) and n i ~ N (0, σn 2) for i = 1, 2, … , N .Based on thisassumption the value of θ can be estimated by taking the numerical average of the nodes observations, i.e.∑==N i i x N 11θ（2）C. Reference System ModelOur reference system consists of N conventional Single Input Single Output (SISO) wireless links, each connecting one of the sensor nodes to the FC. For the reference system we do not consider any communication or cooperation among the sensors. Therefore each sensor quantizes its observation by an L-bit scalar quantizer designed for distribution of θ, generates a message of length L and transmits it directly to the FC. Fusion Center receives all messages and performs the processing, which is calculation of the numerical average of these messages.III. COOPERATIVE DATA PROCESSING ALGORITHMSensor readings are analog quantities. Therefore, each sensor has to compress its data into several bits. For data compression we use L -bit scalar quantizer [12,13].In our algorithm, network is divided into clusters, each cluster having a fixed and pre-defined number of members (m). Members of each cluster are supposed to cooperate with one another in two ways:1. Share, Process and Compress their data2. Cooperatively transmit their processed data using virtual MIMO.IV. ANALYSISThe performance metric considered in our analysis is the total distortion due to compression and errors occurred during transmission. The first distortion is due to finite length quantizer, used in each sensor to represent the analog number by L bits. This distortion depends on the design of quantizer.We consider a Gaussian scalar quantizer which is designed over 105 randomly generated samples. The second distortion is due to errors occurred during transmission through the channel. In our system, this distortion is proportional to the probability of bit error. Since the probability of bit error (Pe) is a function of transmission energy per bit (Eb), total distortion will be a function of Eb. In this section we characterize the transmission and total consumed energy of sensors and find the relationship between distortion and probability of bit error.V. SIMULATION AND NUMERICAL RESULTS To give a numerical example, we assume m = 4 members in each cluster. Therefore our Virtual-MIMO scheme will consist of 4 transmit antennas. We assume that network has N = 32 sensors. Sensor observations are Gaussian with σx2= 1 and are added to a Gaussian noise of σn2= 0.1 .Nodes are deployed uniformly in the field and are 2 meters apart from each other and the Fusion Center is located 100 meters away from the center of the field. The values for circuit parameters are quoted from [6] and are listed in Table I. These parameters depend on the hardware design and technological advances. Fig. 1 illustrates the performance (Distortion) of reference system and proposed two-phase V-MIMO scheme versus transmission energy consumption in logarithmic scale. As shown in the figures, depending on how much precision is needed in the system, we can save energy by applying the proposed algorithm.TABLE IFig. 2 illustrates the Distortion versus total energy consumption of sensor nodes. That is, in this figure we consider both the transmission and circuit energy consumption. The parameters that lead us to these results may be designed to give better performance than presented here. However, from these figures we can conclude that the proposed algorithm outperforms the reference system when we want to have distortion less than 10−3 and it can save energy as high as 10 dB.VI. CONCLUSIONIn this paper we proposed a novel algorithm which takes advantage of cooperation among sensor nodes in two ways: it not only compresses the set of sensor messages at the sensor nodes into one message, appropriate for final estimation but also encodes them into orthogonal space-time symbols which are easy to decode and energy-efficient. This algorithm is able to save energy as high as 10 dB.REFERENCES[1] J.Chou,D.Petrovic and K.Ramchandran “A distributed and adaptive signalprocessing approach to reducing energy consumption in sensornetworks,”Proc. IEEE INFOCOM,March 2003.[2] Z.Q.Luo, “Universal decentralized estimation in a bandwidth constrainedsensor network,” IEEE rmation The ory, vol.51,no.6,June 2005.[3] Z.Q.Luo,“An Isotropic Universal decentralized estimation scheme for abandwidth constrained Ad Hoc sensor network,”IEEEm. vol.23,no. 4,April 2005.[4] Z.Q.Luo and J.-J. Xiao, “Decentralized estimation i n an inhomogeneoussensing environment,” IEEE Trans. Information Theory, vol.51, no.10,October 2005.[5] J.J.Xiao,S.Cui,Z.-Q.Luo and A.J.Goldsmith, “Joint estimation in sensornetworks under energy constraints,” Proc.IEEE First conference on Sensor and Ad Hoc Communications and Networks, (SECON 04),October 2004.[6] S.Cui, A.J.Goldsmith, and A.Bahai,“Energy-efficiency of MIMO andcooperative MIMO techniques in sensor networks,”IEEEm,vol.22, no.6pp.1089–1098,August 2004.[7] S.K.Jayawe era and M.L.Chebolu, “Virtual MIMO and distributed signalprocessing for sensor networks-An integrated approach”,Proc.IEEEInternational Conf. Comm.(ICC 05)May 2005.[8] S.K.Jayaweera,"Energy efficient virtual MIMO-based CooperativeCommunications for Wireless Sensor Networks",2nd International Conf. on Intelligent Sensing and Information Processing (ICISIP 05),January 2005.[9] S.K.Jayaweera,“Energy Analysis of MIMO Techniques in Wireless SensorNetworks”, 38th Annual Conference on Information Sciences and Systems (CISS 04),March 2004.[10] S.K.Jayaweera and M.L.Chebolu,“Virtual MIMO and Distributed SignalProcessing for Sensor Networks - An Integrated Approach”,IEEEInternational Conf.on Communications (ICC 05),May 2005.[11] S.K.Jayaweera,“An Energy-efficient Virtual MIMO CommunicationsArchitecture Based on V-BLAST Processing for Distributed WirelessSensor Networks”,1st IEEE International Conf.on Sensor and Ad-hocCommunications and Networks (SECON 2004), October 2004.[12] J.Max,“Quantizing for minimum distortion,” IRE rmationTheory,vol.IT-6, pp.7 – 12,March 1960.[13] S.P.Lloyd,“Least squares quantization in PCM ,”IEEE rmationTheory,vol.IT-28, pp.129-137,March 1982.[14] D.Slepian and J.K.Wolf “Noiseless encoding of correlated inf ormationsources,” IEEE Trans. on Information Theory,vol.19, pp.471-480,July1973.[15] S.M.Alamouti,“A simple transmit diversity technique for wirelesscommunications,” IEEE m., vol.16,no.8,pp.1451–1458,October 1998.[16] V.Tarokh,H.Jafarkhani,and A.R.Calderbank. “Space-time block codesfrom orthogonal designs,’’IEEE rmationTheory,vol.45,no.5,pp.1456 -1467,July 1999.[17] Y.Oohama,“The Rate-Distortion Function for the Quadratic GaussianCEO Problem,” IEEE Trans. Informatio nTheory,vol.44,pp.1057–1070,May 1998.。

(完整版)外文翻译

(完整版)外文翻译外文文件原稿和译文原稿logistics distribution center location factors:(1)the goods distribution and quantity. This is the distribution center and distribution of the object, such as goods source and the future of distribution, history and current and future forecast and development, etc. Distribution center should as far as possible and producer form in the area and distribution short optimization. The quantity of goods is along with the growth of the size distribution and constant growth. Goods higher growth rate, the more demand distribution center location is reasonable and reducing conveying process unnecessary waste.(2)transportation conditions. The location of logistics distribution center should be close to the transportation hub, and to form the logistics distribution center in the process of a proper nodes. In the conditional, distribution center should be as close to the railway station, port and highway.(3)land conditions. Logistics distribution center covers an area of land in increasingly expensive problem today is more and more important. Is the use of the existing land or land again? Land price? Whether to conform to the requirements of the plan for the government, and so on, in the construction distribution center have considered.(4)commodities flow. Enterprise production of consumer goods as the population shift and change, should according to enterprise's better distribution system positioning. Meanwhile, industrial products market will transfer change, in order to determine the raw materials and semi-finished products of commodities such as change of flow in the location of logistics distribution center should be considered when the flow of the specific conditions of the relevant goods.(5)other factors. Such as labor, transportation and service convenience degree, investment restrictions, etc.(完整版)外文翻译How to reduce logistics cost，enhance the adaptive capacity and strain capacityof distribution center is a key research question of agricultural product logisticsdistribution center．At present，most of the research on logistics cost concentrates offtheoretical analysis of direct factors of logistics cost, and solves the problem ofover-high logistics Cost mainly by direct channel solution ． This research stresseson the view of how to loeate distribution center, analyzes the influence of locatingdistribution center on logistics cost ．and finds one kind of simple and easy locationmethod by carrying on the location analysis of distribution center through computermodeling and the application of Exeel．So the location of agricultural productlogistics distribution center can be achieved scientifically and reasonably, which willattain the goal of reducing logistics cost, and have a decision．making supportfunction to the logisties facilities and planning of agricultural product．The agricultural product logistics distribution center deals with dozens andeven hundreds of clients every day, and transactions are made in high-frequency. Ifthe distribution center is far away from other distribution points，the moving andtransporting of materials and the collecting of operational data is inconvenient andcostly. costly．The modernization of agricultural product logistics s distribution center is acomplex engineering system， not only involves logistics technology, informationtechnology, but also logistics management ideas and its methods，in particular thespecifying of strategic location and business model is essential for the constructing ofdistribution center. How to reduce logistics cost ，enhance the adaptive capacity andstrain capacity of distribution center is a key research question of agricultural productlogistics distribution center. The so— called logistics costs refers to the expendituresummation of manpower, material and financial resources in the moving process of thegoods．such as loading and unloading，conveying，transport，storage，circulating，processing, information processing and other segments. In a word。

外文翻译模板

外文翻译模板(总12页)--本页仅作为文档封面，使用时请直接删除即可----内页可以根据需求调整合适字体及大小--杭州电子科技大学毕业论文外文文献翻译要求根据《普通高等学校本科毕业设计（论文）指导》的内容，特对外文文献翻译提出以下要求：一、翻译的外文文献可以是一篇，也可以是两篇，但总字符要求不少于万（或翻译成中文后至少在3000字以上）。

二、翻译的外文文献应主要选自学术期刊、学术会议的文章、有关着作及其他相关材料，应与毕业论文（设计）主题相关，并作为外文参考文献列入毕业论文（设计）的参考文献。

并在每篇中文译文标题尾部用“脚注”形式注明原文作者及出处，中文译文后应附外文原文（全文，格式为word）。

不能翻译中国学者的文章，不能翻译准则等有译文的着作。

三、中文译文的基本撰写格式1.题目:采用小三号、黑体字、居中打印；段前二行，段后二行。

2.正文:采用小四号、宋体字，行间距一般为固定值20磅，标准字符间距。

页边距为左3cm，右，上下各，页面统一采用A4纸。

四、外文原文格式1.题目:采用小三号、Times New Roman、居中打印；段前二行，段后二行。

2.正文:采用小四号、Times New Roman，行间距一般为固定值20磅，标准字符间距。

页边距为左3cm，右，上下各，页面统一采用A4纸。

五、封面格式由学校统一制作（注：封面上的“翻译题目”指中文译文的题目），并按“封面、封面、译文、外文原文、考核表”的顺序统一装订。

毕业论文外文文献翻译毕业论文题目Xxx翻译题目指翻译后的中文译文的题目学院会计学院（以本模板为准）专业XXXXXX（以本模板为准）姓名XXXXXX（以本模板为准）班级XXXXXX（以本模板为准）学号XXXXXX（以本模板为准）指导教师XXXXXX（以本模板为准）译文管理者过度乐观与债务、股权融资之间的选择1本文采取了一家成长较快速的公司作为样本，比较债务融资和股权融资后的长期股票业绩。

外文文献及翻译(模板)

山东交通学院
2015届毕业生外文文献翻译
院(系)别财经学院
专业
班级
学号
姓名
指导教师
二○一五年六月
China's GEM Will Soon Arrive, but Will It Struggle Like Most
Other High-tech Exchanges
A decade after the initial proposal and following years of anticipation, the launch of China’s Growth Enterprise Market (GEM) is finally within view. The country’s top stock market watchdog -- the China Securities Regulatory Commission (CSRC) -- published rules governing listings on the board at the end of March, and they take effect on May 1. Meanwhile, more rules on auditing committees and sponsors have
also been released.
中国创业板指日可待，它将面临怎样的机会和难题
历经十年的砥砺磨剑，中国创业板指日可待。

中国证券市场的最高监管机构—中国证监会（CSRC）在3月底公布了在创业板上市的规定：《首次公开发行股票并在创业板上市管理暂行办法》，办法将于5月1日起生效。

有关审计委员会和保荐人的规定也已出台。