Abnormal crowd behavior detection by using the particle entropy

The crowd distribution information is the crucial information for abnormal behaviors detection in the crowd scenes. In this paper, we firstly refer to the definition of the entropy and propose an algorithm effectively and accurately representing the crowd distribution information in the crowd scenes. The proposed algorithm not only avoids unstable foreground extraction, but also owns low computational complexity. To detect the abnormal crowd behaviors, we use the Gaussian Mixture Model (GMM) over the normal crowd behaviors to predict the abnormal crowd behaviors since GMM usually can deal well with the unbalanced problem. In this paper we simultaneously use the crowd distribution information and the crowd speed information to estimate the parameters of GMM over the normal crowd behaviors and predict abnormal crowd behaviors. Experiment conducted on publicly available dataset consisting of gathering and dispersion events validates that the proposed approach can preeminently reflect the crowd distribution information. In addition, experiments conducted on publicly UMN dataset demonstrate that the proposed abnormal crowd behavior detection method has an excellent performance and outperforms the state-of-the-art methods.     

A novel approach for crowd video monitoring of subway platforms

Crowd monitoring in a dense crowd scene has become an important and challenging topic in the field of video surveillance system. This paper proposes a novel crowd monitoring approach for subway platforms to address requirements in rail traffic management. Firstly, an improvement for Mixture Gaussian background modeling is presented to segment the crowd. In the process of feature extraction, the concept of the weighted area is proposed to solve the problem of the perspective of images. To deal with the issue of the occlusion between individuals, an improved gradient feature is developed in this paper. And then, Adaptive Boost classifier with the feature weighted area and the improved gradient is used to estimate the crowd density. Finally, the crowd is counted by the method of linear regression. The experimental results show that the proposed approach is feasible and effective for crowd monitoring in real subway platforms.     

Macroscopic effects of microscopic forces between agents in crowd models

Crowd scenarios have attracted attention from computer modellers, perhaps because of the impracticality of studying the phenomenon by traditional experimental methods. For example, Kirchner has proposed an agent-based crowd model inspired by fields of elementary particles [A. Kirchner, A. Schadschneider, Simulation of evacuation processes using a bionics-inspired cellular automaton model for pedestrian dynamics, Physica A 312 (2002) 260–276.], but chose not to incorporate crowd forces. We argue that crowd forces (and associated injuries) are an essential characteristic of crowds, and that their omission will negatively affect the model's ability to make predictions (e.g. time for a crowd to pass through an exit). To support this position we describe an evolution of Kirchner's model that includes a vector-based particle field to represent forces. We show qualitative and quantitative differences compared to Kirchner's model when force is included. The Swarm Force model demonstrates—by showing non-linear effects of force—the necessity of force in crowd models.     

Acoustic classification of Australian anurans based on hybrid spectral-entropy approach

A new hybrid method for automated frog sound identification, using spectral centroid, Shannon entropy and Rényi entropy is proposed. The advantage of using entropy based information theoretic approach for analyzing complexity of bioacoustics signals in animal vocalization is discussed. Sound samples from nine species of Microhylidae frogs are first segmented into syllables. Fourier spectral centroid, Shannon entropy and Rényi entropy of the syllables are then determined. Finally, nonparametric k-th nearest neighbour (k-NN) classifier is used to recognize the frog species based on these three extracted features. Result shows that the k-NN classifier based on these selected features is capable to identify the species of the frogs with an average accuracy of 98%. It is found that the accuracy reduces significantly only when the noise levels higher than −20 dB.     

Discrete element crowd model for pedestrian evacuation through an exit

A series of accidents caused by crowds within the last decades evoked a lot of scientific interest in modeling the movement of pedestrian crowds. Based on the discrete element method, a granular dynamic model, in which the human body is simplified as a self-driven sphere, is proposed to simulate the characteristics of crowd flow through an exit. In this model, the repulsive force among people is considered to have an anisotropic feature, and the physical contact force due to body deformation is quantified by the Hertz contact model. The movement of the human body is simulated by applying the second Newton's law. The crowd flow through an exit at different desired velocities is studied and simulation results indicated that crowd flow exhibits three distinct states, i.e., smooth state, transition state and phase separation state. In the simulation, the clogging phenomenon occurs more easily when the desired velocity is high and the exit may as a result be totally blocked at a desired velocity of 1.6 m/s or above, leading to faster-to-frozen effect.     

Modeling the capability of penetrating a jammed crowd to eliminate freezing transition

Frozen state from jammed state is one of the most interesting aspects produced when simulating the multidirectional pedestrian flow of high density crowds. Cases of real life situations for such a phenomenon are not exhaustively treated.Our observations in the Hajj crowd show that freezing transition does not occur very often. On the contrary, penetrating a jammed crowd is a common aspect. We believe the kindness of pedestrians facing others whose walking is blocked is a main factor in eliminating the frozen state as well as in relieving the jammed state. We refine the social force model by incorporating a new social force to enable the simulated pedestrians to mimic the real behavior observed in the Hajj area.Simulations are performed to validate the work qualitatively.     

基于时频分帧能量熵的陶瓷制品敲击声波信号特征识别

针对现有陶瓷制品敲击声波信号特征提取方法中提取的特征代表性降低的问题,该文提出结合最大重叠离散小波包变换(MODWPT)和时频分帧能量熵的特征提取方法。首先采用MODWPT将信号分解为4层,再对每个节点的子信号分帧后计算各个节点的时频分帧能量熵,然后根据能量分布特征选择了前6个节点的时频分帧能量熵特征,最后构建随机森林分类器完成识别。将该方法和MODWPT时频分段能量熵、MODWPT归一化能量特征两种方法进行比较。实验结果表明,相比MODWPT时频分段能量熵、MODWPT归一化能量两种特征提取方法,MODWPT时频分帧能量熵能提升特征的代表性,具有更优的陶瓷制品敲击声波信号特征识别性能,其识别的F₁值达到了98.46%,相比上述两种方法分别提升F₁值3.22%、1.86%。     

基于径向基函数神经网络的高光谱遥感图像分类

从径向基函数神经网络的理论出发,针对高光谱数据的特点,设计了有效的特征提取模型,再与径向基函数神经网络的输入层连接,建立了一个新的径向基函数神经网络的高光谱遥感影像分类模型,并用国产OMISII传感器获得的64波段数据进行试验。首先进行了最小噪声分离变换,提取了1～20个分量的数据,使用提取后的数据（20维）、提取后数据的纹理变换（20维）和主成分分析的前（20维）,组成了60维向量数据进行分类处理,这种分类器结构简单、容易训练、收敛速度快,其分类精度达到69.27%,高于BP神经网络分类算法(51.20%)以及常用的最小距离分类（MDC）算法(40.88%)。通过对结果和过程进行分析,实验证明径向基函数神经网络在高光谱遥感分类中具有较好的适用性。     

Classification of Indoor Human Fall Events Using Deep Learning

Human fall identification can play a significant role in generating sensor based alarm systems, assisting physical therapists not only to reduce after fall effects but also to save human lives. Usually, elderly people suffer from various kinds of diseases and fall action is a very frequently occurring circumstance at this time for them. In this regard, this paper represents an architecture to classify fall events from others indoor natural activities of human beings. Video frame generator is applied to extract frame from video clips. Initially, a two dimensional convolutional neural network (2DCNN) model is proposed to extract features from video frames. Afterward, gated recurrent unit (GRU) network finds the temporal dependency of human movement. Binary cross-entropy loss function is calculated to update the attributes of the network like weights, learning rate to minimize the losses. Finally, sigmoid classifier is used for binary classification to detect human fall events. Experimental result shows that the proposed model obtains an accuracy of 99%, which outperforms other state-of-the-art models.     

基于高光谱图像技术和SVDD的玉米种子识别

特征提取的充分性和分类器设计的合理性是影响玉米种子识别精度的两个关键问题。采集了玉米种子的高光谱图像,并提取每粒玉米种子在不同波段下的图像熵作为分类特征;在此基础上,利用支持向量数据描述方法构建每类玉米的分类器模型,对待识别样本的测试精度达到了94.14%,对新类别样本的识别精度达到92.28%。仿真结果表明：新方法可实现玉米种子的准确识别,同时解决了传统分类器对新类别样本的错误分类问题。     

基于MCKD-EEMD近似熵和TWSVM的齿轮箱故障诊断

在复杂环境下齿轮箱信号往往会淹没在噪声信号中,特征向量难以提取;为了有效地进行故障诊断,提出了基于最大相关反褶积(MCKD)总体平均经验模态分解(EEMD)近似熵和双子支持向量机(TWSVM)的齿轮箱故障诊断方法;首先采用MCKD方法对强噪声信号进行滤波处理,在采用EEMD方法对齿轮箱信号进行分解,分解后得到本征模函数(IMF)分量进行近似熵求解,得到齿轮特征向量,最后将其输入到TWSVM分类器中进行故障识别;仿真实验表明,采用MCKD-EEMD方法能够有效地提取原始信号,与其他分类器相比,TWSVM的计算时间短,分类效果好等优点。     

Multilevel Deep Feature Generation Framework for Automated Detection of Retinal Abnormalities Using OCT Images

Optical coherence tomography (OCT) images coupled with many learning techniques have been developed to diagnose retinal disorders. This work aims to develop a novel framework for extracting deep features from 18 pre-trained convolutional neural networks (CNN) and to attain high performance using OCT images. In this work, we have developed a new framework for automated detection of retinal disorders using transfer learning. This model consists of three phases: deep fused and multilevel feature extraction, using 18 pre-trained networks and tent maximal pooling, feature selection with ReliefF, and classification using the optimized classifier. The novelty of this proposed framework is the feature generation using widely used CNNs and to select the most suitable features for classification. The extracted features using our proposed intelligent feature extractor are fed to iterative ReliefF (IRF) to automatically select the best feature vector. The quadratic support vector machine (QSVM) is utilized as a classifier in this work. We have developed our model using two public OCT image datasets, and they are named database 1 (DB1) and database 2 (DB2). The proposed framework can attain 97.40% and 100% classification accuracies using the two OCT datasets, DB1 and DB2, respectively. These results illustrate the success of our model.     

仿生鳍条姿态测量及其误差分析

在复杂环境下齿轮箱信号往往会淹没在噪声信号中,特征向量难以提取。为了有效的进行故障诊断,提出了基于最大相关反褶积(MCKD)总体平均经验模态分解(EEMD)近似熵和双子支持向量机(TWSVM)的齿轮箱故障诊断方法。首先采用MCKD方法对强噪声信号进行滤波处理,在采用EEMD方法对齿轮箱信号进行分解,分解后得到本征模函数(IMF)分量进行近似熵求解,得到齿轮特征向量,最后将其输入到TWSVM分类器中进行故障识别。仿真实验表明,采用MCKD-EEMD方法能够有效的提取原始信号,与其他分类器相比,TWSVM的计算时间短,分类效果好等优点。     

Dynamics of emotional contagion in dense pedestrian crowds

In places with high-density pedestrian movements, irrational emotions can quickly spread out under emergency, which may eventually lead to asphyxiation and crushing. It was noticed that a pedestrian's emotion in crowd would change as a result of the influence from other pedestrians. Thus, to explore the dynamics of emotion contagion process in dense pedestrians, two types of pedestrian emotions, i.e., negative and positive have been identified. Taking into account the emotional transit of a pedestrian, a crowd movement model is established in the present paper. We simulate pedestrian movement in a region with periodic boundary condition to study the dynamics of emotional contagion in dense crowds. Influences of the initial negative pedestrian proportion, pedestrian crowd density, emotion influence radius, and dose factor on the transition of overall crowd emotion state have been investigated. We expect this study could provide theoretical suggestions for crowd management.     

Stacked sparse autoencoder in hyperspectral data classification using spectral-spatial,higher order statistics and multifractal spectrum features

This paper proposes a novel classification paradigm for hyperspectral image (HSI) using feature-level fusion and deep learning-based methodologies. Operation is carried out in three main steps. First, during a pre-processing stage, wave atoms are introduced into bilateral filter to smooth HSI, and this strategy can effectively attenuate noise and restore texture information. Meanwhile, high quality spectral-spatial features can be extracted from HSI by taking geometric closeness and photometric similarity among pixels into consideration simultaneously. Second, higher order statistics techniques are firstly introduced into hyperspectral data classification to characterize the phase correlations of spectral curves. Third, multifractal spectrum features are extracted to characterize the singularities and self-similarities of spectra shapes. To this end, a feature-level fusion is applied to the extracted spectral-spatial features along with higher order statistics and multifractal spectrum features. Finally, stacked sparse autoencoder is utilized to learn more abstract and invariant high-level features from the multiple feature sets, and then random forest classifier is employed to perform supervised fine-tuning and classification. Experimental results on two real hyperspectral data sets demonstrate that the proposed method outperforms some traditional alternatives.     

Non-Equilibrium Entropy and Irreversibility in Generalized Stochastic Loewner Evolution from an Information-Theoretic Perspective

In this study, we theoretically investigated a generalized stochastic Loewner evolution (SLE) driven by reversible Langevin dynamics in the context of non-equilibrium statistical mechanics. Using the ability of Loewner evolution, which enables encoding of non-equilibrium systems into equilibrium systems, we formulated the encoding mechanism of the SLE by Gibbs entropy-based information-theoretic approaches to discuss its advantages as a means to better describe non-equilibrium systems. After deriving entropy production and flux for the 2D trajectories of the generalized SLE curves, we reformulated the system’s entropic properties in terms of the Kullback–Leibler (KL) divergence. We demonstrate that this operation leads to alternative expressions of the Jarzynski equality and the second law of thermodynamics, which are consistent with the previously suggested theory of information thermodynamics. The irreversibility of the 2D trajectories is similarly discussed by decomposing the entropy into additive and non-additive parts. We numerically verified the non-equilibrium property of our model by simulating the long-time behavior of the entropic measure suggested by our formulation, referred to as the relative Loewner entropy.     

Robust pedestrian detection in thermal infrared imagery using the wavelet transform

A novel and robust pedestrian detection method in thermal infrared images based on the double-density dual-tree complex wavelet transform (DD-DT CWT) and wavelet entropy is presented in this paper. The regions of interest (ROIs) are located first making use of high brightness property of the pedestrian pixels caused by the self-emission of the pedestrians related to the Planck’s law. The candidate ROIs are then decomposed by DD-DT CWT and the wavelet entropy features are extracted from the high frequency subbands. The true pedestrian regions are finally classified and recognized using the support vector machine (SVM) classifier. Comparisons between our approach and traditional approaches are presented and experimental results using several thermal infrared image databases show the proposed scheme to be very promising.     

Object extraction as a basic process for content-based image retrieval (CBIR) system

This article describes the way in which image is prepared for content-based image retrieval system. Automated image extraction is crucial; especially, if we take into consideration the fact that the feature selection is still a task performed by human domain experts and represents a major stumbling block in the process of creating fully autonomous CBIR systems. Our CBIR system is dedicated to support estate agents. In the database, there are images of houses and bungalows. We put all our efforts into extracting elements from an image and finding their characteristic features in the unsupervised way. Hence, the paper presents segmentation algorithm based on a pixel colour in RGB colour space. Next, it presents the method of object extraction applied to obtain separate objects prepared for the process of introducing them into database and further recognition. Moreover, we present a novel method of texture identification which is based on wavelet transformation. Due to the fact that the majority of texture is geometrical (such as bricks and tiles) we have used the Haar wavelet. After a set of low-level features for all objects is computed, the database is stored with these features.     

Information entropy of humpback whale songs

The structure of humpback whale (Megaptera novaeangliae) songs was examined using information theory techniques. The song is an ordered sequence of individual sound elements separated by gaps of silence. Song samples were converted into sequences of discrete symbols by both human and automated classifiers. This paper analyzes the song structure in these symbol sequences using information entropy estimators and autocorrelation estimators. Both parametric and nonparametric entropy estimators are applied to the symbol sequences representing the songs. The results provide quantitative evidence consistent with the hierarchical structure proposed for these songs by Payne and McVay [Science 173, 587-597 (1971)]. Specifically, this analysis demonstrates that: (1) There is a strong structural constraint, or syntax, in the generation of the songs, and (2) the structural constraints exhibit periodicities with periods of 6-8 and 180-400 units. This implies that no empirical Markov model is capable of representing the songs' structure. The results are robust to the choice of either human or automated song-to-symbol classifiers. In addition, the entropy estimates indicate that the maximum amount of information that could be communicated by the sequence of sounds made is less than 1 bit per second.     

Digital Self-interference cancellation in the era of machine learning: A comprehensive review

In-band full-duplex (IBFD) transmission has been identified as a promising solution to improve spectral efficiency as compared to half-duplex (HD) systems. One of the major challenges of FD systems is to effectively manage the self-interference (SI) produced by the transmitter antenna to the local receiver antenna. In this survey paper, we first review and discuss the preliminaries related to SI cancellation (SIC). Furthermore, this survey presents the advancements to the date of digital SIC approaches and highlights the advantages and limitations of each approach. The survey also summarizes different hardware platforms used for SIC along with their key features. Additionally, we also survey and compare different FD techniques for their SIC performances. Finally, this paper identifies a variety of new research areas in the era of machine learning and deep learning for digital SIC.