查询词典 recognition mark

In order to decrease the error rate of speech recognition, speaker adaptation techniques are often used to adjust speaker-dependent acoustic models. MLLR (Maximum Likelihood Linear Regression) and MAP are two of the most popular techniques in recent years. MLLR uses the technique of regression trees. It calculates the transform matrix for each leaf node of the tree. This makes it possible to use fewer sentences to decrease the error rate of HMM-based speech recognition. However, while we examined the recognition result, we found that although the overall error rate decreased, but the error rate of certain confusable phones was higher. In order to solve this problem, we propose the use MPE (Minimum Phone Error Discriminative Training) to solve this problem.

在语音辨识技术的实际应用上，语者调适技术常被用做调整语者无关之声学模型，使其对於特定语者之辨识率有所提升，常见的语者调适技术有最大化相似度线性回归（Maximum Likelihood Linear Regression），其精神在於透过对语音辨识模型相近的mixture做群聚，再对各个群聚做调整，以期能达到利用少量语料，提升辨识率的目的，但其缺点为，在发音上相近的模型，其mixture原本就极为近似，若再分类为同一群聚做调整，容易因为使用者的发音习惯，而使模型偏向ㄓ或ㄗ，因此虽然整体的辨识率获得提升，但却造成混淆音的错误率上升。

According to the above evaluation of immuno-epidemiology of RSA, RSA may be etiologically classified into lack of materno-fetal immuno-recognition, materno-fetal inappropriate immuno-recognition, and disorder of materno-fetal immuno-recognition.

根据RSA免疫流行病学分析，RSA的病因可以分类为：母-胎免疫识别低下型、母-胎免疫识别过度型及母-胎免疫识别紊乱型。

The experimental results show that the SOM method can get the best result for lesion recognition, the recognition rate of pork tissue and liver tissue are 96.1% and 91.7%, respectively. The recognition rate for PCA-LDA method is lower than that of SOM method because the feature of tissue lesion is linear indivisible.An effective clinic scheme is developed for NETE, and some problems are discussed. For example, nonuniform temperature field conduced by nonuniform energy field; tissue database construction; movement of HIFU transducer with ultrasonic transducer; rectification of HIFU parameters.

临床方案做了总体设计，并讨论了在临床应用中需要考虑的一些问题，如因HIFU辐射能量分布不均匀导致的焦域的温度场不均匀的问题，数据库的建立问题，HIFU换能器和超声探头的位移的问题以及HIFU辐射参数的确定问题等。

The history and current situation of research in speech recognition are proposed all-sidedly and systematically. We probe into such topics about acoustic features, phonetic features, phonologic features and auditory perception in terms of related theory and disciplines. Furthermore, we summarize and analyze focus in current theoretical and technical aspects of speech recognition and immediately put an insight into challenge to speech recognition presently, even in a long time.

论文全面、系统地介绍了国内外语音识别的发展历史和发展现状，结合相关学科的知识，分析了语音声学特征、语音学特征、音系学特征及其语音感知等有关问题，分析并总结了目前语音识别理论和技术实现方面的研究重点以及目前、甚至在相当长的时期内语音识别面临的挑战。

Therefore, it is very convenient to construct photoactive supramolecular recognition systems, which fulfil the requirements of optical chemical sensors: to complete two affairs simultaneously, a recognition between host and guest, b signal expression related to recognition.

利用超分子体系的特性，可以方便地构造满足光化学传感器要求的具有光学活性的超分子识别体系：在识别分析对象的过程中，必须给出与识别过程相关的光学参数来完成信号表达过程。

A character-codes model based pattern recognition method by using amplitude data of targets frequency response is proposed after the establishment of high frequency response equation of radar targets. The method uses the same unvariant character-codes to express the same target of different angles and uses sequential separating of regions of target's reference variables to obtain multiple coding vectors of all regions for every target. The category and pose of target can be determined after a two-staged classification processing such as coding processing of target range profile data and errors correcting processing of coded data. The two-stage classification results are used to target identification as posterior hypothesis to be tested by sufficient convincing and necessary convincing. The character-codes model based pattern recognition method is also adaptive to variation of target angles and controllable in error probability. Time-consuming and complicated iterative computation is not necessary in the train process of pattern recognition.

首先建立了雷达目标的高频频率响应方程，这种方程将目标多散射中心理论与目标局部谐振理论有机地结合起来；进而提出了目标多频响应结构成像方法，这种方法将瑞利区频率响应或谐振区频率响应曲线的不同部位用各种不同的近似方法去逼近，通过特征谱估计方法将目标的局部频响特征与宽带距离响应特征结合起来；然后提出了一种基于姿态角变化区域的序贯划分与多模匹配的特征编码识别方法，这种方法具有能适应目标姿态角的变化以及能控制识别的差错概率α等优点，在目标识别时只需要目标在少数有限个频率激励下的幅频响应数据。

Firstly, this thesis lucubrates in the method theory of character space and does a lot of experiments and researches in the two steps, character expressing and measure of similarity. Realize the object recognition system based on the character subspace, which adopts the effective and ameliorative means to solve the problems existing in the course of recognition such as illumination and pose estimation, so enhances the robustness of recognition.

首先深入研究特征空间方法原理，并对识别过程中的两个关键步骤：特征表达和相似性度量做了具体实验和研究工作，实现基于特征空间目标识别系统，其中采用了几种有效的改进方法，有效的解决了识别过程中存在的诸如光照以及3D目标姿态估计等问题，提高了识别鲁棒性。

As a bionic sense technique, the AITS is composed of three parts, namely the cross-sensitive sensors and their array, self-study specialist database and intelligent recognition. The voltammetric and the potentiometric techniques are two main test methods. The sensitive materials include noble metals, rare earth metals, lipids and lipids polymer, chalcogenide glass materials, conductive polymers and other materials such as graphite paste, phthalocyanin, Prussia blue etc. The PCA and the ANN are the primary pattern recognition tools in intelligent recognition.

作为一种仿生感觉技术，人工智能味觉系统是由交互感应电极及阵列、自学习专家数据库和智能模式识别三部分构成；电位法和伏安法是两类主要的传感方法；传感器材料包括贵金属及稀土金属、类脂/聚合物、硫属玻璃、导电聚合物以及其它如碳糊、酞菁、普鲁士蓝等；主成分分析和人工神经网络是智能识别过程中主要的模式识别工具。

Considering the speed and performa nce of the recognition system, Cyberglove is selected as gesture input device in sign language recognition system, DGMM is used as recognition technique, and hierarchical recognizer is used in recognizing mo dule, which can recognize 274 sign language words coming from the dictionary of Chinese sign language with an accuracy of 97.4%, based on Chinese sign language' s own characteristic.

考虑到系统的实时性及识别效率，该系统选取Cyberglove型号数据手套作为手语输入设备，采用DGMM(dynamic Gaussian mixt ure model)作为系统的识别技术，并根据中国手语的具体特点，在识别模块中选取了多层识别器，可识别中国手语字典中的274个词条，识别率为97.4%。

The credibility evaluation index system and feedback correction decision-making mechanism of the recognition result were established by qualitative and quantitative analysis of three kinds of general characters recognition error. And the analytical method of the recognition error was given.

设计了3种广义识字误差，通过对这3种广义识字误差的类型和数值进行定性与定量相结合的分析，建立了识别结果的可信度评价指标体系和反馈校正决策机制。

In the chapter 2, the theoretic knowledge about the photosensitive resin and the grinding tools was firstly introduced.

第二章阐述了光固化树脂结合剂磨具的相关理论研究。

Do not use the program's indenting or margin-setting features; these will be added during typesetting.

不要使用缩排，页面边缘设置之类的选项，偶看不大懂。