查询词典 figure of speech

The word-to-sound rule and the structure of a speech database are modified during the implementation of TTS system, and then a filter algorithm based on part-of-speech is proposed to solve the polyphone problem. Prosodic research on Chinese characters plays a great role in speech recognition and speech synthesis system. In order to improve the naturalness of a speech synthesis system, the relationship among the sentence structure, the length of a sentence and the breath boundaries is analyzed and a breath boundary assigning algorithm based on such relationship is put forward. In addition, the author analyzes the process of a speech synthesis system and builds a prosodic markup language suitable for Chinese speech synthesis.

呼吸群边界划分作为汉语韵律的一种表现方式，它在语音感知和文语转换系统的性能方面起着十分重要的作用，作者在对汉语句法结构、句子长度和和呼吸群划分之间关系的研究基础上，设计并实现了基于句子长度和句法结构的呼吸群边界划分算法；为更好地描述文语转换过程，作者提出了适合汉语语音合成的SSML韵律标注语言，该标注语言不仅从声学层上对语音的韵律进行描述，而且从语言学层上标注句法结构信息，最后本文对基于语言理解的韵律规则用SSML进行了描述并用实验验证了语言学层面的标注对系统自然度的影响。

G. speech perception and understanding. These stages are described and interpreted in terms of the findings from recorded corpus and questionnaires in education. As for speech production, students' speech motives are explored in detail. Speaker's speech intention is interpreted in connection with listener's reception of it. Findings are presented to show that, in the course of formulating speech plan, the previous sentences is often processed, the construction of sentence is a process of accumulative expansion of speech items through sequential or inversive appending or insertion, and the interaction between long-term and short-term speech plans often occurs.

本文运用录音语料、问卷调查材料具体分析了师生在言语生成、言语理解各阶段的特点和规律：在言语生成方面，本文探讨了学生的言语动机及其分类；结合言语意图的领会考察了言语意图；发现在制订言语计划过程中，常常出现对所听话语的处理加工，语句的建构是一个语项累加扩展过程，这种累加可以是顺序的，也可以是插序或倒序的，言语计划的制订还表现为预定计划与即时计划的相互作用；考察了言语计划执行过程中的一些现象，如制订计划与执行计划的交替、执行计划的中断等，并分析了产生这些现象的原因。

Figure 1 water tank, Figure 2 cabinets, Figure 3 connector Figure 4 vacuum tube, Figure5 health-care formulations stone, Figure 6 protective plate Figure 7 fastened rooting angle iron, Figure 8 booster pump, Figure9 filter.

附图1是水箱，2是橱柜，3是连接器，4是真空管，5保健配方石，6是防护板，7是紧固生根角铁，8是增压泵，9是过滤器。

Some information such as pitch and harmonic of each speech stream, interaural time difference and interaural intensity difference and etc could be obtained from overlapping speech by the analyzing and modeling of overlapping speech and noisy speech, based on which the interested speech can be reconstructed and overlapping speech can be separated.

通过对混叠语音和含噪语音进行分析建模，对混叠语音分别提取其各自语音流的基音及谐波、双耳时间差ITD和强度差IID等信息，并以此重构目标语音分量，达到分离混叠语音的目的。

Hidden Image is an image contains one or more embedded figures The hidden figure merges into typically a busy image such that it does not standout and remains unnoticeable for a while It takes the viewer to focus or some targeted effort to detect the hidden figure and might requires some hint to be able to do so In this paper we provide the technological processes to create the hidden image which normal Texture Synthesis can't be done; According to quantization of background and figure we manage to recolor the quantized color of figure through Luminance Labeling Processing which is base on relationship between background and figure we also manage to remove some parts assigned by user to make our result more natural but can still recognize the figure Finally we present the result using the Texture Processing We believe the proposed system can provide a user-friendly tool to create their own Hidden Image

Hidden Image，中文称为画中有画的艺术；在表现手法上是一们相当新颖且前卫的艺术，其表现风格多种；一般来说，常见手法是将一个或是多个物体以不同型态的方式巧妙的融合在背景当中，观察者在远方观赏整幅画时，并不会觉得特别突兀，而当视觉集中到某一个焦点上时，透过大脑重新组织之后，其隐藏的物体将渐渐浮现出来。本篇论文主要是提供一个完整的流程；根据luminance space区块化后，透过物体本身区块的相互关系，或是物体与背景的相互关系进行Luminance Labeling，并经由Erosion Processing将物体做适当的破坏，最后以贴图方式给予视觉化呈现。相信透过这样的流程，我们将赋予一般普罗大众一套机制去创造出属於自己的Hidden Image。

This dissertation begins with the analysis of speech signal pretreatment and does some points test, draws the characteristic parameter of the introduction pronunciation through MATLAB speech processing toolbox, gets rid of the aphonia segment and yawp of speech signal, offers effective speech segment for the distill of speech characteristic parameter.

本系统首先从语音信号的预处理开始分析，通过MATLAB语音处理工具箱提取输入语音的特征参数，对语音信号进行了端点检测，滤除了语音信号的无声段和噪声，为语音特征参数的提取提供了有效语音段。

For the pitch misgrouping problem caused by no voicing decision in Meddis'pitch extracting method for concurrent speech signals, referring to voicing decision for single speech signal, basing on the analysis of the voicing charater of concurrent speech signals, we present a voicing decision method in pitch extracting for concurrent speech signals. Double periodic degree and frequencies group's distribution degree are introduced for voicing decision of concurrent speech signals.

针对Meddis的基于听觉心理生理的混叠语音基频提取算法中不分清浊音所导致的语音基频错判问题，借鉴了单语音清浊音判断的方法，基于对混叠语音的清浊性能的分析，提出了混叠语音基频提取中的清浊音判断算法，提出了判断混叠语音周期性的双周期度及判断频带组周期性的分布度两个指标，用于混叠语音基频提取中的清浊音判断。

Main on-going research works include:"Expressive speech synthesis system","Spontaneous Speech Generation in Natural Human Computer Interaction"(863),"High performance distributed speech synthesis and recognition engine"(863),"The evaluation system of Chinese information processing"(863),"Multilingual speech synthesis system","emotional speech synthesis and classification","text independent voice conversion", etc.

当前正在进行的研究工作包括：具有表现力的语音合成系统，自然人机交互中的口语产生系统（863），高质量分布式语音合成和语音识别引擎的研究（863），中文信息处理评价系统（863），多语言语音合成系统，情感语音合成与情感分类，文本独立的语音转换系统等。

It summarizes the current advancements in the aboveOmentioned field in terms of eight aspects :①relations between L2 speech perception and production ,②L2 phonological studies ,③L2 speech segmental studies ,④L2 speech acquisition models ,⑤L2 speech suprasegmental studies ,⑥L1 transfer in L2phonological acquisition ,⑦nonO linguistic factors in L2 phonological acquisition ,and ⑧L2 speech instruction.

本文主要探讨了国外第二语言语音习得研究近年来的进展，着重涉及以下8 个方面：①二语口头话语感知与表达的关系、②二语口头话语的音系学研究、③二语口头话语的语段理论研究、④二语音系理论模型的构建、⑤二语口头话语的超音段研究、⑥二语音系习得中的母语迁移、⑦非语言因素对二语音系习得的影响、⑧二语语音教学研究。

Tenthly, this Chinese should be smart and eloquence as obama, remind you, when choose an assistant, he should not find the one with Chinese stylel who draft out all speech lecture, as those people can make this simplest issues into most complicated ones, making a 10 minutes speech to 30 minutes, woshington is the model he should learn in the aspect of speech , on march 4,1793, Washington gave a 135 words of speech during inaugural speech .

第十，这位华裔本人一定要像奥巴马一样聪明绝顶，雄辩超群。需要提醒他的是，在选择助手时，他一定不能找中国式的秘书为他起草任何演讲稿，因为那些人常常把最简单的问题极端复杂化，使你把本来可以用十分钟结束的演讲放大到四小时三十分钟。这位华裔在演讲方面应该学习的榜样是华盛顿，1793年3月4日，华盛顿在他的总统就任典礼上发表了一篇只有135字的就职演说。

You can snipe the second and third union leaders from this position.

您可以鹬第二和第三工会领袖从这一立场出发。

Aiming at the currently shortage of XML streams quality detecting, this paper proposes a new forecasting method of XML streams quality by least squares support vector machines, which is used the method of XML keys' vector matrix as windows, and vector product wavelet transform to multilevel decompose and refactor the XML streams series, that can fulfill real-time checking demand of XML quality, and ensure constraint, consist- ency and integrality. For even more adapting net load, it proposes a control strategy by weight and adaptive adjustment to ensure XML streams quality.

针对当前XML数据流质量检测存在的不足，提出构建XML键的矢量矩阵作为窗口，利用矢量积小波变换多级分解与重构XML数据流，再结合最小二乘支持向量机对XML数据流质量进行预测的一种方法，满足XML数据流质量重构时实时检测的要求，保证XML数据的约束性、一致性与完整性；为了更好的适应网络负载，采取加权与自适应窗口调整等调度策略充分保证XML数据流的质量检测。