TONE MODELING BASED ON HIDDEN CONDITIONAL RANDOM FIELDS AND DISCRIMINATIVE MODEL WEIGHT TRAINING

The use of hidden conditional random fields （HCRFs） for tone modeling is explored. The tone recognition performance is improved using HCRFs by taking advantage of intra-syllable dynamic, inter-syllable dynamic and duration features. When the tone model is integrated into continuous speech recognition, the discriminative model weight training （DMWT） is proposed. Acoustic and tone scores are scaled by model weights discriminatively trained by the minimum phone error （MPE） criterion. Two schemes of weight training are evaluated and a smoothing technique is used to make training robust to overtraining problem. Experiments show that the accuracies of tone recognition and large vocabulary continuous speech recognition （LVCSR） can be improved by the HCRFs based tone model. Compared with the global weight scheme, continuous speech recognition can be improved by the discriminative trained weight combinations.

Adaptive foreground and shadow segmentation using hidden conditional random fields

Video object segmentation is important for video surveillance, object tracking, video object recognition and video editing. An adaptive video segmentation algorithm based on hidden conditional random fields (HCRFs) is proposed, which models spatio-temporal constraints of video sequence. In order to improve the segmentation quality, the weights of spatio-temporal con- straints are adaptively updated by on-line learning for HCRFs. Shadows are the factors affecting segmentation quality. To separate foreground objects from the shadows they cast, linear transform for Gaussian distribution of the background is adopted to model the shadow. The experimental results demonstrated that the error ratio of our algorithm is reduced by 23% and 19% respectively, compared with the Gaussian mixture model (GMM) and spatio-temporal Markov random fields (MRFs).

Prediction of Potential Disease-Associated MicroRNAs Based on Hidden Conditional Random Field

MicroRNAs( miRNAs) are reported to be associated with various diseases. The identification of disease-related miRNAs would be beneficial to the disease diagnosis and prognosis. However,in contrast with the widely available expression profiling, the limited knowledge of molecular function restrict the development of previous methods based on network similarity measure. To construct reliable training data,the decision fusion method is used to prioritize the results of existing methods. After that,the performance of decision fusion method is validated. Furthermore,in consideration of the long range dependencies of successive expression values,Hidden Conditional Random Field model( HCRF) is selected and applied to miRNA expression profiling to infer disease-associated miRNAs. The results show that HCRF achieves superior performance and outperforms the previous methods. The results also demonstrate the power of using expression profiling for discovering disease-associated miRNAs.

PREDICTION OF LOCATION OF HIDDEN ORE DEPOSITS IN THE AGED ORE FIELDS: AN EXAMPLE FROM FENGHUANGSHAN ORE FIELD, TONGLING, CHINA

This paper discusses the strategy for successfully predicting the location of potential hidden ore bodies in aged ore field, and presents the result of location prediction of hidden ore bodies in Fenghuangshan ore field, Tongling. Innovative conceptual targeting procedures based on a genetic understanding of mineralization systems, carefully geological investigation and correct deduction, together with new geochemical and geophysical technology and integrating of comprehensive information are all very important for the successful prediction. In the aged Fenghuangshan ore field, through researching by application of the metallogenic theory of polygenetic compound ore deposits and triple frequency induced polarization method and exploration tectono geochemical method, we predicted location and quality of hidden ore bodies. According to the prediction, hidden high quality Cu Au ore bodies of skarn type and porphyry type have been discovered.

煤田震电协同勘探技术研究与应用

近年来煤田震电协同勘探技术在设计、处理分析方面都取得了较大的发展,从采区最常使用的三维地震勘探、瞬变电磁勘探方法入手,对采集到的各种数据运用主成分分析(PCA)进行融合处理,解译煤矿采区工作面的异常地质体及富水性等隐蔽致灾因素。该技术实现了三维地震、瞬变电磁勘探在数据层面的综合勘探,提高了物探解释煤矿采区工作面地质异常体(富水性)的准确性,为震电协同勘探技术体系的构建提供了新思路,为煤矿安全、高效、绿色生产提供了保障。

烃气组分研究与隐伏矿预测——以箭猪坡矿区为例

烃气具有扩散速度快、运移距离远、携带信息量大等特点,较常规化探方法优势明显。通过五圩矿田箭猪坡矿体与围岩、矿床外围南北边两条化探剖面土壤样品采集与测试分析,研究矿体、围岩烃气组分特征,认为箭猪坡矿床成因与岩浆热液密切相关;对比土壤烃气异常点与不同地质体、土壤烃气背景值的烃气组分标准化曲线特征,筛选与成矿热液密切相关的烃气异常,认为矿床南北两侧的I、II、III号有利勘查区内烃气异常与岩浆热液密切相关,推测形成三排洞、水落、箭猪坡的成矿热液在该区内亦有活动,存在伏矿体,已有钻孔于II号有利区内钻遇厚5cm含闪锌矿包及黄铜矿的石英脉,IV号有利勘查区内烃气异常与变质热液相关,且已发现有汞矿点,推测有隐伏矿体存在。

Malus’ Law Derived from Deterministic Particle Behavior

A polarized beam of energy is usually interpreted as a set of particles, all having the same polarization state. Difference in behavior between the one and the other particle is then explained by a number of counter-intuitive quantum mechanical concepts like probability distribution, superposition, entanglement and quantized spin. Alternatively, I propose that a polarized beam is composed of a set of particles with a cosine distribution of polarization angles within a polarization area. I show that Malus’ law for the intensity of a beam of polarized light can be derived in a straightforward manner from this distribution. I then show that none of the above-mentioned counter-intuitive concepts are necessary to explain particle behavior and that the ontology of particles, passing through a polarizer, can be easily and intuitively understood. I conclude by formulating some questions for follow-up research.

多维GMM与邻域约束的多光谱机载LiDAR数据城市地物分类

如何充分利用空间位置及多光谱信息完整、准确地区分各类地物是多光谱机载激光雷达点云应用的重要前提。传统的基于点的分类算法受点云无法明晰表达拓扑及邻域信息的局限导致算法设计困难、执行效率低,而将点云插值为图像的分类算法则受图像存在信息及精度损失的局限导致分类精度较低。另外,点云和图像两种结构均无法直接表达地物的三维几何形体,不利于地物三维建模及分析。为此,本文提出了一种多维高斯混合模型(Gauss mixture model,GMM)与邻域约束的多光谱机载LiDAR城区地物分类算法。该算法首先以无损且明晰表达邻域信息为原则将多光谱LiDAR数据体素化为多值虚拟体素结构,其中,虚拟体素为体素与其内激光点的联合体,虚拟体素值是由体素内激光点的多波段光谱、高程、局部法向量分布及点密度等特征构成的特征矢量。然后,构建模糊聚类模型对多值虚拟体素结构进行分割,获得各虚拟体素的模糊隶属度矩阵。其中,特征空间地物呈现的多峰分布用多维GMM拟合,从而建立标号场,并将多维GMM的概率分布作为非相似性测度;标号场中相邻体素类别的空间相关性用隐马尔可夫随机场模型建模,从而建立邻域约束下的先验概率,并将其作为控制聚类尺度的参数;采用附有规则化项的目标函数解决聚类尺度敏感问题。最后,对隶属度矩阵进行反模糊化确定分类结果。采用Optech Titan实测的不同场景的、不同数据量的多光谱机载LiDAR数据定量评价本文算法的有效性和可行性。试验结果表明,本文算法的平均总体精度可达91.32%、Kappa系数可达0.872,可有效实现对城市各类地物的分类;本文算法综合利用了地物的辐射、空间及几何一致性信息,扩大了信息利用种类,为多光谱机载LiDAR数据的空间位置及多光谱信息的综合利用提供了可行方案。

Mathematical Expression Extraction in Text Fields of Documents Based on HMM

Aiming at the problem that the mathematical expressions in unstructured text fields of documents are hard to be extracted automatically, rapidly and effectively, a method based on Hidden Markov Model (HMM) is proposed. Firstly, this method trained the HMM model through employing the symbol combination features of mathematical expressions. Then, some preprocessing works such as removing labels and filtering words were carried out. Finally, the preprocessed text was converted into an observation sequence as the input of the HMM model to determine which is the mathematical expression and extracts it. The experimental results show that the proposed method can effectively extract the mathematical expressions from the text fields of documents, and also has the relatively high accuracy rate and recall rate.

基于HMM+CRF词性标注的实体抽取方法

基于HMM+CRF词性标注的实体抽取方法从词性标注入手,对待处理文本先进行词性标注;然后根据文本的词性将实体抽取出来,在传统的CRF词性标注模型上增加一层HMM模型,提高实体抽取的精确度;最后在人民日报语料上进行实验,验证了准确率分别在基于HMM实体抽取模型和基于CRF实体抽取模型的基础上提高了2.1%和0.3%。

岩质边坡微震监测方案优化设计及隐伏危险区域辨识

地压监测是矿山安全管理的关键环节,也是矿产资源开发过程中有效的监测手段。根据陕西省四方金矿的矿山地质条件和地压活动规律,为提高资源回收率,有效治理矿山地表环形边坡,解决安全绿色采矿的关键问题,针对矿山生产特点,结合传感器布设的基本要求,分析了不同布设方案的理想定位误差,并从经济性角度分析与设计了满足矿山监测要求的地压系统监测方案。为进一步辨识矿山边坡体的薄弱异常区域,确定边坡的区域稳定性情况,采用波速场成像技术获取了边坡体监测剖面低速危险区域和高速危险区域的成像结果,进而实现了边坡体内部隐伏空洞异常区域的超前辨识及监测效果分析。结果表明:在理想化条件下,某一监测点设置为不同的标高时,坡体定位误差差距不明显,结合监测震源的矿山边坡区域相对波速场成像结果可知,基于当前的台网布置方案,可满足对于矿山边坡体监测剖面内部危险区域的辨识需求。研究成果对于减少矿山灾害事故发生、提升矿山安全生产稳定性具有重要意义。

基于多模型融合的电力运检命名实体识别

为有效解决构建电力运检知识图谱的关键步骤之一的电力运检命名实体识别问题,通过构建一种基于Stacking多模型融合的隐马尔可夫-条件随机场-双向长短期记忆网络(hidden Markov-conditional random fields-bi-directional long short-term,HCB)模型方法研究了电力运检命名实体识别问题。HCB模型分为两层,第一层使用隐马尔可夫模型(hidden Markov model,HMM)、条件随机场(conditional random fields,CRF)和双向长短期记忆网络(bi-directional long short-term memory,Bi-LSTM)模型进行训练预测,再将预测结果输入第二层的CRF模型进行训练,经过双层模型训练预测得出最后的命名实体。结果表明:在电力运检命名实体识别问题上HCB模型的精确率、召回率及F1值等指标明显优于单模型以及其他的融合模型。可见HCB模型能有效解决电力运检命名实体识别问题。

云南省澜沧老厂地区重力场特征及其地质意义

根据2006年全国危机矿山(澜沧县澜沧铅矿)接替资源勘查和2017年云南省地勘基金矿产地质调查项目部署的重力测量成果,经过数据处理,结合地层(岩石)密度特征、地质、矿床等资料综合分析研究,推测了老厂及外围地区隐伏中酸性岩体空间形态特征;研究分析了深部隐伏夕卡岩型铜钼矿体的重力特征,提出了老厂及其外围寻找深部隐伏矿床的区域,对老厂及其外围矿产勘查具有较明确的指导意义.

在寻找金属矿山深部隐伏矿中地气测量的应用

为了探讨在寻找金属矿山深部隐伏矿中地气测量的应用,采用了理论与实践相结合的方法。首先基于地气测量的原理和技术特点进行分析,其次分析了地气测量在寻找金属矿山深部隐伏矿中的应用要点、存在的问题和发展趋势。研究表明,寻找金属矿山深部隐伏矿具有很强的技术性和复杂性,影响因素较多,找矿难度巨大。传统的找矿方法存在很大的局限性,准确性不足,且工作量大。然而,地气测量技术是一种新型的找矿技术,能够通过分析地气来发现隐伏矿的位置、深度和种类,从而为隐伏矿的开采利用提供有效的数据支撑,促进我国矿产资源事业的持续健康发展。

Ontology of Relativistic Mass

The term “relativistic mass” defined by equation m=γm0 with γ=(1-v2/c2)-1/2 has a somewhat controversial history, based on special relativity theory, mathematics, logic, intuition, experiment, and ontology. Key is the ontological framework, specifically whether the framework does or does not include gravity. This paper examines both cases, with detailed analysis of gravitomagnetism and of relativistic mass in collisions.

基于隐条件随机场的自适应视频分割算法

视频目标分割是视频监视与视频目标跟踪、视频目标识别以及视频编辑的基础.本文提出了一种基于隐条件随机场(Hidden conditional random fields,HCRF)的自适应视频分割算法,利用HCRF模型对视频序列中的时空邻域关系建模.使用在线学习的方式对相应的参数进行调整,实现对时空邻域约束关系的权重调整,提高视频目标分割细节上的效果.大量的数据测试表明,与高斯混合模型(Gaussian mixture model,GMM)和联合时空的马尔可夫随机场(Markov random fields,MRF)等算法相比,该算法的分割错误率分别降低了23%和19%.

基于ECC和HFE的纠错密码构造

纠错密码是一种利用纠错码体制来实现纠错和加密双重功能的一种密码体制。大部分已知的纠错密码从变换的角度看是一种对明文的线性变换。从密码分析的角度看,由于不具有非线性变换,密码的混淆能力不强,容易被攻击。利用纠错码(Error-Correction Code,ECC)改造基本HFE(Hidden Field Equations)密码算法,所得的新密码算法具有纠错和加密功能,而且因其具有概率密码特性以及建立在MQ困难问题之上,具有很高的安全强度。

基于轮廓特征的人体行为识别

提出了一种以人的动作序列图像的轮廓为特征、基于隐条件随机场的行为识别方法。首先,利用背景差分法和阴影消除技术提取运动人体轮廓。星型骨架方法只采用单一质心-边界距离来描述人体轮廓,因此对轮廓的局部特征描述能力较弱。定义一种新的基于距离组的轮廓描述方法,它将时变的2D轮廓形状转换为对应的1D距离向量。最后利用判别隐条件随机场对行为进行训练和识别。实验结果表明,本方法的正确识别率达到91.4%以上,识别结果较为理想。

结合HMRF模型的模糊ISODATA高分辨率遥感图像分割

模糊ISODATA(Fuzzy ISODATA,FISODATA)在应用于图像分割时,FISODATA算法定义的FCM目标函数未考虑邻域像素间的数据相关性,导致该算法的抗噪性能较差;此外,FISODATA算法中分裂-合并操作需人工选取阈值参数,而不适当的阈值往往使得该算法陷入局部极值,因而得到错误的类属数并影响图像分割结果。该文将考虑邻域关系的基于隐马尔可夫随机场(Hidden Markov Random Field,HMRF)FCM(HMRF-FCM)方法纳入ISODATA框架,提出HMRF-FCM ISODATA(HMRF-FISODATA)算法,在分裂与合并操作后增加了优化操作,并根据优化结果自适应调节控制聚类分裂与合并的各阈值。该算法不仅能够快速获取正确类属数,而且克服了FISODATA算法没有考虑邻域像素的关系、人工选取阈值参数和受图像噪声影响大等问题,实现了自动正确确定类属数的同时完成高精度图像分割。

基于贪心EM算法的HMRF遥感影像变化检测

提出一种基于贪心EM算法的HMRF遥感影像变化检测算法。该算法采取PCA与差值法相结合的方式来构造差分影像。首先,采用隐马尔可夫随机场(Hidden Markov Random Field,HMRF)模型描述空间上下文信息,并构造系统能量函数;然后,利用贪心EM算法克服EM算法假定混合成分数为已知、迭代结果过分依赖初始值、可能收敛到局部最大点或收敛到参数空间边界的缺点,能够准确学习分布模型结构和参数,发现数据对模型的最佳匹配;最后,通过条件迭代模型(Iterated Conditional Modes,ICM)优化算法求解能量函数最优解,获取变化区域。实验结果表明,该算法能够更好地保持影像的结构性,有效去除孤立噪声。