Spatial correlations in time and frequency domains between chlorophyll-a concentration and environmental factors in the Bohai Sea

Based on the reconstructed MODIS data and ECMWF reanalysis data from 2003 to 2021,spatial correlations between chlorophyll a(Chl a)and sea surface temperature(SST),photosynthetically available radiation(PAR),aerosol optical thickness(AOT),and wind speed(WS)in the Bohai Sea were analyzed from the perspective of time domain and frequency domain.Results indicate that the frequency domain analysis was more conducive to revealing the correlations between Chl a and environmental factors.The spatial pattern of time-domain correlations was similar to the isobaths of the Bohai Sea,which was positive in shallow waters and negative in deep waters for SST,PAR,and AOT,and was reversed for WS.Frequency-domain correlations were obtained by performing Fourier Transform and were higher than correlations in time domain.The spatial distributions indicated that the effects of SST and PAR on Chl a were greater than AOT and WS in the Bohai Sea.Additionally,cross-spectrum analysis was applied to explore the response relationships.A depth-dependent pattern was shown in correlations and time lags,indicating that the influential mechanism of environmental factors on Chl-a concentration is related to seawater depth.

Correlation Analysis of X-Band Sea Clutter in Complex Domain

The correlation analysis of sea clutter data in a complex domain is conducted in this study. Specific to X-band sea clutter, the statistical characteristics of the complex correlation, particularly the phase characteristics which are closely related to the phase difference of the sea clutter and the Doppler properties, are analyzed in detail based on the experimental data, recorded by the Mc Master University IPIX radar in 1993. That the phase term of the complex correlation presents linear change means that there exists the linearity of phase differences between different time intervals in the X-band sea clutter. This investigation explores the regularities about the effect of wind on the complex correlation with similar patterns for different polarization modes. The regularities indicate that the wind direction can be inferred from the distribution pattern of the complex correlation. Moreover, a model describing the relationships between the statistics of the complex correlation and wind parameters is proposed. The application for target detection based on the differences of characteristics of complex correlations between the sea clutter and the target are also investigated and the proposed features have been confirmed. The principle of the method is fundamental for broader future applications.

SPEECH ENHANCEMENT BASED ON DYNAMIC NOISE ESTIMATION WITHIN AUTO-CORRELATION DOMAIN

Most noise suppression algorithms of single channel use the mean of noisy segments to estimate the characteristics of noise spectrum, ignoring the estimation of noise in speech segments. Therefore, when the energy level of noise varies with the time, the performance of removing noise will be degraded. To solve this problem, a speech enhancement approach based on dynamic noise estimation within correlation domain was proposed. This method exploits the characteristics that noise energy mainly concentrates on 0 th order correlation coefficients, signal is auto correlated but signal and noise, noise and noise are uncorrelated, then estimates and decomposes the noise, thus helps to solve the above mentioned problem. The results of recognition experiments on speech signals of 15 Chinese cities’ names corrupted by noise of exhibition hall shows, this approach is better than SS (Spectral Subtraction) method, adapts better to the variances of energy levels of speech signal corrupted by noise, has some practicability to improve the robustness of recognition systems under noisy environment.

3D Gravity Inversion with Correlation Image in Space Domain and Application to the Northern Sinai Peninsula

We present a 3D inversion method to recover density distribution from gravity data in space domain.Our method firstly employs 3D correlation image of the vertical gradient of gravity data as a starting model to generate a higher resolution image for inversion.The 3D density distribution is then obtained by inverting the correlation image of gravity data to fit the observed data based on classical inversion method of the steepest descent method.We also perform the effective equivalent storage and subdomain techniques in the starting model calculation,the forward modeling and the inversion procedures,which allow fast computation in space domain with reducing memory consumption but maintaining accuracy.The efficiency and stability of our method is demonstrated on two sets of synthetic data and one set of the Northern Sinai Peninsula gravity data.The inverted 3D density distributions show that high density bodies beneath Risan Aniza and low density bodies exist to the southeast of Risan Aniza at depths between 1~10 and 20 km,which may be originated from hot anomalies in the lower crust.The results show that our inversion method is useful for 3D quantitative interpretation.

MODAL PARAMETERS EXTRACTION WITH CROSS CORRELATION FUNCTION AND CROSS POWER SPECTRUM UNDER UNKNOWN EXCITATION

In most of real operational conditions only response data are measurable while the actual excitations are unknown, so modal parameter must be extracted only from responses. This paper gives a theoretical formulation for the cross-correlation functions and cross-power spectra between the outputs under the assumption of white-noise excitation. It widens the field of modal analysis under ambient excitation because many classical methods by impulse response functions or frequency response functions can be used easily for modal analysis under unknown excitation. The Polyreference Complex Exponential method and Eigensystem Realization Algorithm using cross-correlation functions in time domain and Orthogonal Polynomial method using cross-power spectra in frequency domain are applied to a steel frame to extract modal parameters under operational conditions. The modal properties of the steel frame from these three methods are compared with those from frequency response functions analysis. The results show that the modal analysis method using cross-correlation functions or cross-power spectra presented in this paper can extract modal parameters efficiently under unknown excitation.

A WYNER-ZIV VIDEO CODING METHOD UTILIZING MIXTURE CORRELATION NOISE MODEL

In Wyner-Ziv (WZ) Distributed Video Coding (DVC), correlation noise model is often used to describe the error distribution between WZ frame and the side information. The accuracy of the model can influence the performance of the video coder directly. A mixture correlation noise model in Discrete Cosine Transform (DCT) domain for WZ video coding is established in this paper. Different correlation noise estimation method is used for direct current and alternating current coefficients. Parameter estimation method based on expectation maximization algorithm is used to estimate the Laplace distribution center of direct current frequency band and Mixture Laplace-Uniform Distribution Model (MLUDM) is established for alternating current coefficients. Experimental results suggest that the proposed mixture correlation noise model can describe the heavy tail and sudden change of the noise accurately at high rate and make significant improvement on the coding efficiency compared with the noise model presented by DIStributed COding for Video sERvices (DISCOVER).

Boundary Correlation Functions of the gl(1|1) Supersymmetric Vertex Model

The gl（1/1） supersymmetric vertex model with domain wall boundary conditions （DWBC） on an N × N square lattice is considered. We derive the reduction formulae for the one-point boundary correlation functions of the model. The determinant representation for the boundary correlation functions is also obtained.

3D prestack reverse time migration of ground penetrating radar data based on the normalized correlation imaging condition

The reverse time migration(RTM)of ground penetrating radar(GPR)is usually implemented in its two-dimensional(2D)form,due to huge computational cost.However,2D RTM algorithm is difficult to focus the scattering signal and produce a high precision subsurface image when the object is buried in a complicated subsurface environment.To better handle the multi-off set GPR data,we propose a three-dimensional(3D)prestack RTM algorithm.The high-order fi nite diff erence time domian(FDTD)method,with the accuracy of eighth-order in space and second-order in time,is applied to simulate the forward and backward extrapolation electromagnetic fi elds.In addition,we use the normalized correlation imaging condition to obtain pre-stack RTM result and the Laplace fi lter to suppress the low frequency noise generated during the correlation process.The numerical test of 3D simulated GPR data demonstrated that 3D RTM image shows excellent coincidence with the true model.Compared with 2D RTM image,the 3D RTM image can more clearly and accurately refl ect the 3D spatial distribution of the target,and the resolution of the imaging results is far better.Furthermore,the application of observed GPR data further validates the eff ectiveness of the proposed 3D GPR RTM algorithm,and its fi nal image can more reliably guide the subsequent interpretation.

Bearing Fault Diagnosis Based on Deep Discriminative Adversarial Domain Adaptation Neural Networks

Intelligent diagnosis driven by big data for mechanical fault is an important means to ensure the safe operation ofequipment. In these methods, deep learning-based machinery fault diagnosis approaches have received increasingattention and achieved some results. It might lead to insufficient performance for using transfer learning alone andcause misclassification of target samples for domain bias when building deep models to learn domain-invariantfeatures. To address the above problems, a deep discriminative adversarial domain adaptation neural networkfor the bearing fault diagnosis model is proposed (DDADAN). In this method, the raw vibration data are firstlyconverted into frequency domain data by Fast Fourier Transform, and an improved deep convolutional neuralnetwork with wide first-layer kernels is used as a feature extractor to extract deep fault features. Then, domaininvariant features are learned from the fault data with correlation alignment-based domain adversarial training.Furthermore, to enhance the discriminative property of features, discriminative feature learning is embeddedinto this network to make the features compact, as well as separable between classes within the class. Finally, theperformance and anti-noise capability of the proposedmethod are evaluated using two sets of bearing fault datasets.The results demonstrate that the proposed method is capable of handling domain offset caused by differentworkingconditions and maintaining more than 97.53% accuracy on various transfer tasks. Furthermore, the proposedmethod can achieve high diagnostic accuracy under varying noise levels.

Direction-of-arrival estimation based on direct data domain (D3) method

A direction-of-arrival （DOA） estimation algorithm based on direct data domain （D3） approach is presented. This method can accuracy estimate DOA using one snapshot modified data, called the temporal and spatial two-dimensional vector reconstruction （TSR） method. The key idea is to apply the D3 approach which can extract the signal of given frequency but null out other frequency signals in temporal domain. Then the spatial vector reconstruction processing is used to estimate the angle of the spatial coherent signal source based on extract signal data. Compared with the common temporal and spatial processing approach, the TSR method has a lower computational load, higher real-time performance, robustness and angular accuracy of DOA. The proposed algorithm can be directly applied to the phased array radar of coherent pulses. Simulation results demonstrate the performance of the proposed technique.

TIME DOMAIN PROCESSING MODE SPREAD SPECTRUM SYSTEM

The construction and specifications of a surface acoustic wave storage correlator are described. A time domain processing mode spread spectrum system is presented. An analysis of the interference rejection for this system is provided. The formula for calculating the probability of error of the system is given. The experimental results agree with the theoretical analysis.

Prediction of film ratings based on domain adaptive transfer learning

This paper examines the prediction of film ratings.Firstly,in the data feature engineering,feature construction is performed based on the original features of the film dataset.Secondly,the clustering algorithm is utilized to remove singular film samples,and feature selections are carried out.When solving the problem that film samples of the target domain are unlabelled,it is impossible to train a model and address the inconsistency in the feature dimension for film samples from the source domain.Therefore,the domain adaptive transfer learning model combined with dimensionality reduction algorithms is adopted in this paper.At the same time,in order to reduce the prediction error of models,the stacking ensemble learning model for regression is also used.Finally,through comparative experiments,the effectiveness of the proposed method is verified,which proves to be better predicting film ratings in the target domain.

An enriched environment increases the expression of fibronectin type Ⅲ domain-containing protein 5 and brain-derived neurotrophic factor in the cerebral cortex of the ischemic mouse brain

Many studies have shown that fibronectin type III domain-containing protein 5(FDNC5) and brain-derived neurotrophic factor(BDNF) play vital roles in plasticity after brain injury. An enriched environment refers to an environment that provides animals with multi-sensory stimulation and movement opportunities. An enriched environment has been shown to promote the regeneration of nerve cells, synapses, and blood vessels in the animal brain after cerebral ischemia;however, the exact mechanisms have not been clarified. This study aimed to determine whether an enriched environment could improve neurobehavioral functions after the experimental inducement of cerebral ischemia and whether neurobehavioral outcomes were associated with the expression of FDNC5 and BDNF. This study established ischemic mouse models using permanent middle cerebral artery occlusion(pMCAO) on the left side. On postoperative day 1, the mice were randomly assigned to either enriched environment or standard housing condition groups. Mice in the standard housing condition group were housed and fed under standard conditions. Mice in the enriched environment group were housed in a large cage, containing various toys, and fed with a standard diet. Sham-operated mice received the same procedure, but without artery occlusion, and were housed and fed under standard conditions. On postoperative days 7 and 14, a beam-walking test was used to assess coordination, balance, and spatial learning. On postoperative days 16–20, a Morris water maze test was used to assess spatial learning and memory. On postoperative day 15, the expression levels of FDNC5 and BDNF proteins in the ipsilateral cerebral cortex were analyzed by western blot assay. The results showed that compared with the standard housing condition group, the motor balance and coordination functions(based on beam-walking test scores 7 and 14 days after operation), spatial learning abilities(based on the spatial learning scores from the Morris water maze test 16–19 days after operation), and memory abilities(based on the memory scores of the Morris water maze test 20 days after operation) of the enriched environment group improved significantly. In addition, the expression levels of FDNC5 and BDNF proteins in the ipsilateral cerebral cortex increased in the enriched environment group compared with those in the standard housing condition group. Furthermore, the Pearson correlation coefficient showed that neurobehavioral functions were positively associated with the expression levels of FDNC5 and BDNF(r = 0.587 and r = 0.840, respectively). These findings suggest that an enriched environment upregulates FDNC5 protein expression in the ipsilateral cerebral cortex after cerebral ischemia, which then activates BDNF protein expression, improving neurological function. BDNF protein expression was positively correlated with improved neurological function. The experimental protocols were approved by the Institutional Animal Care and Use Committee of Fudan University, China(approval Nos. 20160858 A232, 20160860 A234) on February 24, 2016.

一种结构不确定分析的改进多维平行六面体模型

多维平行六面体模型是一种能够同时考虑相关变量和独立变量的非概率凸集模型,更适用于工程结构中常见的多源不确定性问题。为更加有效合理地度量结构不确定性,提出一种改进多维平行六面体模型。通过定义区间变量的相关角和边缘区间,给出模型不确定域的显式表达式,进而给出依据实验样本点构建多维平行六面体模型的方法。3个算例分析结果表明,改进多维平行六面体模型能够较好地反映区间变量之间的相关性,是一种比传统多维平行六面体模型更为紧凑合理的模型。

Analysis of the Frequency Domain Water-level Deconvolution Method Based on Reservoir Airgun Source Data

Based on the 2016 airgun experimental data of the Fujian Nanyi reservoir,we adopted the frequency domain water-level deconvolution method and cross-correlation time delay detection technique to study the influence of level scaling factor and the background noise level of the station on deconvolution calculation results, and analyze the effect of deconvolution on eliminating the influence of the source caused by different air-gun pressures. The results show that:( 1) When the level scaling factor is smaller,the signal to noise ratio of the waveform after the deconvolution is smaller,and when the level scaling factor is over smaller,the identification error of travel time is greater.( 2) When the SNR of the station record is higher,the recognition accuracy of travel time is higher,the influence of SNR on the reference station record is far greater than the far station,when the SNR of the far station record is more than 10,the error of travel time is within6 ms,but when the SNR of the reference station record is 30,the travel time error may reach to 20 ms.( 3) When the airgun source difference is big,the frequency domain waterlevel deconvolution method has better effect on eliminating the source influence,but the method error may be introduced when the source difference is small.

基于敏感特征深度域关联的Android恶意应用检测方法

利用机器学习或深度学习算法进行Android恶意应用的检测是当前主流方法,取得了一定的效果。然而,多数方法仅关注应用的权限和敏感行为等信息,缺乏对敏感行为协同的深度分析,导致恶意应用检测准确率低。对敏感行为协同深度分析的挑战主要有两个:表征敏感特征域关联和基于敏感特征域关联的深层分析与检测。本文提出了一种新的Android恶意应用检测模型GCNDroid,基于敏感特征域关联关系图描述的应用程序主要敏感行为以及敏感行为之间的域关联关系来有效地检测Android恶意应用。首先,为了筛选出对分类更加敏感的特征,同时减少图节点的数量,加速分析,本文构建了敏感特征字典。接着,定义类或者包为域,在同一个域中的敏感特征具有域关联关系。通过敏感特征所在域的相对范围,构造敏感特征之间不同的域关联权重,生成敏感特征域关联关系图,敏感特征域关联关系图可以准确表征特定功能模块中的敏感行为,以及敏感行为之间的完整关系。然后,基于敏感特征域关联关系图,设计基于图卷积神经网络的深度表征,构建Android恶意应用检测模型GCNDroid。在实践中,GCNDroid还可以利用新的敏感特征不断更新,以适应移动应用程序新的敏感行为。最后,本文对GCDNroid进行了系统评估,召回率、调和平均数、AUC等重要指标均超过96%。与传统的机器学习算法(支持向量机和决策树)和深度学习算法(深度神经网络和卷积神经网络)相比,GCNDroid取得了预期的效果。

主题方面共享的领域主题层次模型

层次主题模型是构建主题层次的重要工具.现有的层次主题模型大多通过在主题模型中引入nCRP构造方法,为文档主题提供树形结构的先验分布,但无法生成具有明确领域涵义的主题层次结构,即领域主题层次.同时,领域主题不仅存在层次关系,而且不同父主题下的子主题之间还存在子领域方面共享的关联关系,在现有主题关系研究中没有合适的模型来生成这种领域主题层次.为了从领域文本中自动、有效地挖掘出领域主题的层次关系和关联关系,在4个方面进行创新研究.首先,通过主题共享机制改进nCRP构造方法,提出nCRP+层次构造方法,为主题模型中的主题提供具有分层主题方面共享的树形先验分布;其次,结合nCRP+和HDP模型构建重分层的Dirichlet过程,提出rHDP(reallocated hierarchical Dirichlet processes)层次主题模型;第三,结合领域分类信息、词语语义和主题词的领域代表性,定义领域知识,包括基于投票机制的领域隶属度、词语与领域主题的语义相关度和层次化的主题-词语贡献度;最后,通过领域知识改进rHDP主题模型中领域主题和主题词的分配过程,提出结合领域知识的层次主题模型rHDP_DK(rHDP with domain knowledge),并改进采样过程.实验结果表明,基于nCRP+的层次主题模型在评价指标方面均优于基于nCRP的层次主题模型(hLDA,nHDP)和神经主题模型(TSNTM);通过rHDP_DK模型生成的主题层次结构具有领域主题层次清晰、关联子主题的主题词领域差异明确的特点.此外,该模型将为领域主题层次提供一个通用的自动挖掘框架.

基于FDS的飞机客舱火灾数值模拟计算域优化

为了提高飞机客舱火灾数值模拟的精确度,采用火灾动力学三维模拟软件(fire dynamic simulation, FDS)建立全尺寸A320客舱火灾模型,分析了可燃物热解导致的计算域变化对温度特征的影响,通过引入相关性分析方法重点研究了不同计算区域条件下飞机客舱内部热流场变化情况。结果表明,考虑燃烧过程中可燃物的热解变化会增加客舱热释放速率和改变温度场,最大差值可达328℃。计算区域延展尺寸大小将影响到飞机客舱内部的热流场计算,当计算区域扩展到2 m时,客舱温度场和火源处温度场的相关性值分别增大至0.975 0和0.964 2,为了在保证数值模拟计算精度的前提下尽量减少计算时间,建议将计算区域向外延展2 m。

深度语义关联学习的基于图像视觉数据跨域检索

基于图像的视觉数据跨域检索任务旨在搜索与输入图像在语义上一致或外形上相似的跨域图像和三维模型数据,其面临的主要问题是处理跨域数据之间的模态异质性。现有方法通过构建公共特征空间,采用域适应算法或深度度量学习算法实现跨域特征的域对齐或语义对齐,其有效性仅在单一类型的跨域检索任务中进行了验证。提出一种基于深度语义关联学习的方法,以适用多种类型的基于图像的跨域视觉数据检索任务。首先,使用异构网络提取跨域数据的初始视觉特征;然后,通过构建公共特征空间实现初始特征映射,以便进行后续的域对齐和语义对齐;最后,通过域内鉴别性学习、域间一致性学习和跨域相关性学习,消除跨域数据特征之间的异质性,探索跨域数据特征之间的语义相关性,并为检索任务生成鲁棒且统一的特征表示。实验结果表明,该方法在TU-Berlin、IM2MN和MI3DOR数据集中的平均精度均值(mAP)分别达到0.448、0.689和0.874,明显优于对比方法。