Skip to main content
main-content

Über dieses Buch

This two-volume set LNCS 9225 and LNCS 9226 constitutes - in conjunction with the volume LNAI 9227 - the refereed proceedings of the 11th International Conference on Intelligent Computing, ICIC 2015, held in Fuzhou, China, in August 2015. The 191 full papers and 42 short papers of this volume were carefully reviewed and selected from 191 submissions. The papers are organized in topical sections such as evolutionary computation and learning; compressed sensing, sparse coding and social computing; neural networks, nature inspired computing and optimization; pattern recognition and signal processing; image processing; biomedical informatics theory and methods; differential evolution, particle swarm optimization and niche technology; intelligent computing and knowledge discovery and data mining; soft computing and machine learning; computational biology, protein structure and function prediction; genetic algorithms; artificial bee colony algorithms; swarm intelligence and optimization; social computing; information security; virtual reality and human-computer interaction; healthcare informatics theory and methods; unsupervised learning; collective intelligence; intelligent computing in robotics; intelligent computing in communication networks; intelligent control and automation; intelligent data analysis and prediction; gene expression array analysis; gene regulation modeling and analysis; protein-protein interaction prediction; biology inspired computing and optimization; analysis and visualization of large biological data sets; motif detection; biomarker discovery; modeling; simulation; and optimization of biological systems; biomedical data modeling and mining; intelligent computing in biomedical signal/image analysis; intelligent computing in brain imaging; neuroinformatics; cheminformatics; intelligent computing in computational biology; computational genomics; special session on biomedical data integration and mining in the era of big data; special session on big data analytics; special session on artificial intelligence for ambient assisted living; and special session on swarm intelligence with discrete dynamics.

Inhaltsverzeichnis

Frontmatter

Parallel Collaborative Filtering Recommendation Model Based on Two-Phase Similarity

Problems such as cold startup, accuracy, and scalability are faced by traditional collaborative filtering recommendation algorithm if the system is expanded continuously. To resolve these issues, we propose a parallel collaborative filtering recommendation model on the basis of two-phase similarity (PCF-TPS) and weighted distance similarity measure (WDSM). In accordance with WDSM, the users’ similarity is calculated and their similarity matrix is obtained. At the same time, the items’ similarity is counted and its similarity matrix is got in line with Tanimoto Coefficient Similarity. For the users’ similarity matrix, their preferences are endowed with weights and in this way their new preferences matrix is received. In addition, the nearest neighbor item is found and a more accurate recommendation to the target user is given on the basis of the items’ similarity matrix and users’ new preferences matrix. Besides, in regard to the parallel computing framework, the parallel implementation of the model is completed. All these experiments are done on MovieLens dataset. The results show that PCF-TPS solves the problem of cold startup and increases the accuracy concerning CF. Compared with PCF-EV, PCF-TPS’s parallel realization can be improved to nearly 125 times on the whole. That is to say, it will be more meaningful to complex model using GPU than a small model. What’s more, PCF-EV’s distributed implementation is much more efficient than PCF-EV’s.

Hongyi Su, Xianfei Lin, Caiqun Wang, Bo Yan, Hong Zheng

An Improved Ant Colony Algorithm to Solve Vehicle Routing Problem with Time Windows

This paper presents an improved ant colony optimization algorithm (ACO algorithm) based on Ito differential equations, the proposed algorithm integrates the versatility of Ito thought with the accuracy of ACO algorithm in solving the vehicle routing problem (VRP), and it executes simultaneous move and wave process, and employs exercise ability to unify move and wave intensity. Move and wave operator rely on attractors and random perturbations to set the motion direction. In the experiment part, this improved algorithm is implemented for solving vehicle routing problem with soft time windows (VRPSTW), and tested by Solomon Benchmark standard test dataset, the result shows that the proposed algorithm is effective and feasible.

Yi Yunfei, Lin Xiaodong, Sheng Kang, Cai Yongle

Evolutionary Nonnegative Matrix Factorization for Data Compression

This paper aims at improving non-negative matrix factorization (NMF) to facilitate data compression. An evolutionary updating strategy is proposed to solve the NMF problem iteratively based on three sets of updating rules including multiplicative, firefly and survival of the fittest rules. For data compression application, the quality of the factorized matrices can be evaluated by measurements such as sparsity, orthogonality and factorization error to assess compression quality in terms of storage space consumption, redundancy in data matrix and data approximation accuracy. Thus, the fitness score function that drives the evolving procedure is designed as a composite score that takes into account all these measurements. A hybrid initialization scheme is performed to improve the rate of convergence, allowing multiple initial candidates generated by different types of NMF initialization approaches. Effectiveness of the proposed method is demonstrated using Yale and ORL image datasets.

Liyun Gong, Tingting Mu, John Y. Goulermas

Learning-Based Evolutionary Optimization for Optimal Power Flow

This paper proposes a learning-based evolutionary optimization (LBEO) for solving optimal power flow (OPF) problem. The LBEO is a simple and effective algorithm, which simplifies the structure of teaching-learning-based optimization (TLBO) and enhances the convergence speed. The performance of this method is implemented on IEEE 30-bus test system with the minimized fuel cost objective function, and the results show that LBEO is practicable for OPF problem compared with other methods in the literature.

Qun Niu, Wenjun Peng, Letian Zhang

Blind Nonparametric Determined and Underdetermined Signal Extraction Algorithm for Dependent Source Mixtures

Blind extraction or separation statistically independent source signals from linear mixtures have been well studied in the last two decades by searching for local extrema of certain objective functions, such as nonGaussianity (NG) measure. Blind source extraction (BSE) algorithm from underdetermined linear mixtures of the statistically dependent source signals is derived using nonparametric NG measure in this paper. After showing that maximization of the NG measure can also separate or extract the statistically weak dependent source signals, the nonparametric NG measure is defined by statistical distances between different source signals distributions based on the cumulative density function (CDF) instead of traditional probability density function (PDF), which can be estimated by the quantiles and order statistics using the

$$ L^{2} $$

norm efficiently. The nonparametric NG measure can be optimized by a deflation procedure to extract or separate the dependent source signals. Simulation results for synthesis and real world data show that the proposed nonparametric extraction algorithm can extract the dependent signals and yield ideal performance.

Fasong Wang, Rui Li, Zhongyong Wang, Xiangchuan Gao

The Chaotic Measurement Matrix for Compressed Sensing

How to construct a measurement matrix with good performance and easy hardware implementation is the core research problem in compressed sensing. In this paper, we present a simple and efficient measurement matrix named Incoherence Rotated Chaotic (IRC) matrix. We take advantage of the well pseudorandom of chaotic sequence, introduce the concept of the incoherence factor and rotation, and adopt QR decomposition to obtain the IRC measurement matrix which is suited for sparse reconstruction. Simulation results demonstrate IRC matrix has a better performance than Gaussian random matrix, Bernoulli random matrix and other state-of-the-art measurement matrices. Thus it can efficiently work on both natural image and remote sensing image.

Shihong Yao, Tao Wang, Weiming Shen, Shaoming Pan, Yanwen Chong

Leveraging Semantic Labeling for Question Matching to Facilitate Question-Answer Archive Reuse

A new question representation method is proposed for automated question matching over accumulated question-answer data archive. The representation defines four kinds of question words as question-type words, user-centered words, shareable-pattern words, and irrelevant words for question analysis. These question words are further annotated by a semantic labeling ontology to enhance the semantic representation for the purpose of word ambiguity reduction. We tested the matching precision on 5,000 questions with respect to various generators and the result demonstrated the stability of the method. We further compared the method with Cosine similarity and WordNet-based semantic similarity as baselines on a standard TREC dataset containing 5,536 questions. The results presented that our method improved MRR by 8.6 % and accuracy by 9.6 % on average, indicating its effectiveness.

Tianyong Hao, Xinying Qiu, Shengyi Jiang

How to Detect Communities in Large Networks

Community detection is a very popular research topic in network science nowadays. Various categories of community detection algorithms have been proposed, such as graph partitioning, hierarchical clustering, partitional clustering. Due to the high computational complexity of those algorithms, it is impossible to apply those algorithms to large networks. In order to solve the problem, Blondel introduced a new greedy approach named lovian to apply to large networks. But the remained problem lies in that the community detection result is not unstable due to the random choice of seed nodes. In this paper, we present a new modularity optimization method, LPR, for community detection, which chooses the node in order of the PageRank value rather than randomly. The experiments are executed by using medium-sized networks and large networks respectively for community detection. Comparing with lovian algorithm, the LPR method achieves better performance and higher computational efficiency, indicating the order of choosing seed nodes greatly influences the efficiency of community detection. In addition, we can get the importance values of nodes which not only is part of our algorithm, but also can be used to detect the community kernel in the network independently.

Yasong Jiang, Yuan Huang, Peng Li, Shengxiang Gao, Yan Zhang, Yonghong Yan

Self-adaptive Percolation Behavior Water Cycle Algorithm

Water cycle algorithm is a new meta-heuristic optimization algorithm based on the observation of water cycle and how rivers and streams flow downhill towards the sea in the real world. In this paper, a new self-adaptive water cycle algorithm with percolation behavior is proposed. The percolation behavior is introduced to accelerate the convergence speed of proposed algorithm. At the same time, a self-adaptive rainfall process can generate the new stream, more and more new position can be explored, consequently, increasing the diversity of population. Eight typical benchmark functions are tested, the simulation results show that the proposed algorithm is feasible and effective than basic water cycle algorithm, and demonstrate that this proposed algorithm has superior approximation capabilities in high-dimensional space.

Shilei Qiao, Yongquan Zhou, Rui Wang, Yuxiang Zhou

Oscillatory Behavior in An Inertial Six-Neuron Network Model with Delays

This paper discusses the existence of oscillatory solutions in an inertial six neurons BAM neural network model with delays. By means of Chafee’s criterion of limit cycle, some sufficient conditions to ensure the existence of oscillatory solutions for this delayed system are provided. Computer simulations verify the correctness of the results.

Chunhua Feng, Zhenkun Huang

An Online Supervised Learning Algorithm Based on Nonlinear Spike Train Kernels

The online learning algorithm is shown to be more appropriate and effective for the processing of spatiotemporal information, but very little researches have been achieved in developing online learning approaches for spiking neural networks. This paper presents an online supervised learning algorithm based on nonlinear spike train kernels to process the spatiotemporal information, which is more biological interpretability. The main idea adopts online learning algorithm and selects a suitable kernel function. At first, the Laplacian kernel function is selected, however, in some ways, the spike trains expressed by the simple kernel function are linear in the postsynaptic neuron. Then this paper uses nonlinear functions to transform the spike train model and presents the detail experimental analysis. The proposed learning algorithm is evaluated by the learning of spike trains, and the experimental results show that the online nonlinear spike train kernels own a super-duper learning effect.

Xianghong Lin, Ning Zhang, Xiangwen Wang

Forecasting Weather Signals Using a Polychronous Spiking Neural Network

Due to its inherently complex and chaotic nature predicting various weather phenomena over non trivial periods of time is extremely difficult. In this paper, we consider the ability of an emerging class of temporally encoded neural network to address the challenge of weather forecasting. The Polychronous Spiking Neural Network (PSNN) uses axonal delay to encode temporal information into the network in order to make predictions about weather signals. The performance of this network is benchmarked against the Multi-Layer Perceptron network as well as Linear Predictor. The results indicate that the inherent characteristics of the Polychronous Spiking Network make it well suited to the processing and prediction of complex weather signals.

David Reid, Hissam Tawfik, Abir Jaafar Hussain, Haya Al-Askar

A Water Wave Optimization Algorithm with Variable Population Size and Comprehensive Learning

Water wave optimization (WWO) is a new nature-inspired metaheuristic by mimicking shallow water wave motions including propagation, refraction, and breaking. In this paper we present a variation of WWO, named VC-WWO, which adopts a variable population size to accelerate the search process, and develops a comprehensive learning mechanism in the refraction operator to make stationary waves learn from more exemplars to increase the solution diversity, and thus provides a much better tradeoff between exploration and exploitation. Experimental results show that the overall performance of VC-WWO is better than the original WWO and other comparative algorithms on the CEC 2015 single-objective optimization test problems, which validates the effectiveness of the two new strategies proposed in the paper.

Bei Zhang, Min-Xia Zhang, Jie-Feng Zhang, Yu-Jun Zheng

Water Wave Optimization for the Traveling Salesman Problem

Water wave optimization (WWO) is a novel evolutionary algorithm borrowing ideas from shallow water wave models for global optimization problems. This paper presents a first study on WWO for a combinatorial optimization problem — the traveling salesman problem (TSP). We adapt the operators in the original WWO so as to effectively exploring in a discrete solution space. The results of simulation experiments on a set of test instances from TSPLIB show that the proposed WWO algorithm is not only applicable and efficient for TSP, but also has significant performance advantage in comparison with two other methods, genetic algorithm (GA) and biogeography-based optimization (BBO).

Xiao-Bei Wu, Jie Liao, Zhi-Cheng Wang

Improved Dendritic Cell Algorithm with False Positives and False Negatives Adjustable

In order to overcome the blindness of the evaluation on contexts in the classical Dendritic Cell Algorithm (DCA), how weight matrixes influence detection results is analyzed, and two kinds of DCA which can adjust false positives and false negatives are proposed. The first one is the improved voting DCA, the Tendency Factor (TF) is involved in the Dendritic Cell (DC) state transition to assess contexts fairly, and through the fine adjustment of TF false positives and false negatives of detection results are controlled; the other one is the scoring DCA, in the DC state transition phase the evaluation of contexts is ignored, instead, the antigen is directly given a score, then according to the distribution of average scores of antigens the anomaly threshold value can be adjusted to control false positives and false negatives. Experiments show that the two algorithms can both effectively realize results controlled, comparatively the scoring DCA is more intuitive.

Song Yuan, Xin Xu

Topographic Modulations of Neural Oscillations in Spiking Networks

We present a computational model evoked by electrosensory system which is able to display oscillatory activity, and focus on the coherence of the spectral power of the ELL neurons with the topographic modulations for different spatial scale regimes. Numerical simulations reveal that the spatial scale is a very important determinant of neural oscillations in gamma band. The spectral power is enhanced by decreasing feedback spatial spread. This enhancement can also occur if the feedforward is global. However, when the feedforward is topographic, the oscillations saturate to a steady state. In brief, the topographic feedback alone enables the system to modulate gamma activity with the spatial scale, while the introduction of topography in feedforward brings little effect on oscillations. What our results further indicate is that the topographic feedback can induce and enhance oscillations even when the external stimulus is local.

Jinli Xie, Jianyu Zhao, Qinjun Zhao

Non-negative Approximation with Thresholding for Cortical Visual Representation

This paper presents a neurally plausible algorithm for the representation of visual inputs by cortical neurons. It has been demonstrated in previous theoretical studies that the main goal of the encoding of the input from lateral geniculate nucleus (LGN) by simple cell is to minimize the representation error. Based on the existing methods, we propose a non-negative approximation algorithm using thresholding. We validate the algorithm via simulation of several known response properties of simple cells, including the sharp and contrast invariant orientation tuning and surround suppression, and as cross orientation suppression.

Jiqian Liu, Chunli Song, Chengbin Zeng

Targeting the Minimum Vertex Set Problem with an Enhanced Genetic Algorithm Improved with Local Search Strategies

The minimum feedback vertex set in a directed graph is a

NP

-hard problem i.e., it is very unlikely that a polynomial algorithm can be found to solve any instances of it. Solutions of the minimum feedback vertex set can find several real world applications. For this reason, it is useful to investigate heuristics that might give near-optimal solutions. Here we present an enhanced genetic algorithm with an ad hoc local search improvement strategy that finds good solutions for any given instance. To prove the effectiveness of the algorithm, we provide an implementation tested against a large variety of test cases. The results we obtain are compared to the results obtained by greedy and randomized algorithms for finding approximate solutions to the problem.

Vincenzo Cutello, Francesco Pappalardo

Lie Detection from Speech Analysis Based on K–SVD Deep Belief Network Model

Considering the task of lie detection relates some nonlinear characteristics, such as psychological acoustics and auditory perception, which are difficult to be extracted and have high computational complexity. So this paper proposes a deep belief network based on the K-singular value decomposition (K-SVD) algorithm. This method combined the multi-dimensional data linear decomposition ability of sparse algorithm and the deep nonlinear network structure of deep belief network. It is aim to extract the significant time dynamic deep lie structure characteristics. Based on these deep characteristics, the lie database of Arizona University at United States was used to test. The experimental results show that, compared with the K-SVD sparse characteristics and basic acoustic characteristics, the deep characteristics proposed in this paper has better recognition rate. Furthermore, this paper provides a new exploration for psychology calculation.

Yan Zhou, Heming Zhao, Xinyu Pan

Automatic Seizure Detection in EEG Based on Sparse Representation and Wavelet Transform

Sparse representation has been widely applied to pattern classification in recent years. In the framework of sparse representation based classification (SRC), the test sample is represented as a sparse linear combination of the training samples. Due to the epileptic EEG signals are non-stationary and transitory, wavelet transform as a time-frequency analysis method is widely used to analyze EEG signals. In this work, a novel EEG signal classification method based on sparse representation and wavelet transform was proposed to detect the epileptic EEG from EEG recordings. The frequency subbands decomposed by wavelet transform provided more information than the entire EEG. The experimental results showed that the proposed method could classify the ictal EEG and interictal EEG with accuracy of 98 %.

Shanshan Chen, Qingfang Meng, Yuehui Chen, Dong Wang

Design of Serial Communication Module Based on Solar-Blind UV Communication System

UV communication is a wireless optical communication technology which bases on the scattering and absorption of UV light in atmospheric. It combines the advantages of traditional optical communications and wireless communications, including non-line-of-sight (NLOS), anti-interference, low wiretapping and electromagnetic silent. With a wide range of serial communications for remote monitoring and control, the demand for serial communications is increasing in engineering applications. This paper presents the application of serial port in non line-of-sight UV communications to meet the requirements of low-rate, short-range communication. A serial communication interface based on Altera Company’s Cyclone II series FPGA chip (EP2C35F672C6) has been designed. The Universal Asynchronous Receiver Transmitter (UART) module is generated by the Verilog HDL programming language in the integrated software development environment (Quartus II 9.0). Simulation and debugging results indicate that it shows good functions and satisfies protocol requirements.

Chunpei Li, Xiangdong Luo, Huajian Wang

Target Tracking via Incorporating Multi-modal Features

The challenge of visual tracking is to develop a robust target’s appearance model, the core of which involves an appropriate selection and an effective assembly of a cluster of features. In this paper, we propose a novel model to adaptively choose reasonable combination of feature sets to represent the target by employing the multi-kernel ridge regression. This model will update the weights distributions of different kernel groups for feature sets automatically and the regularization parameter’s value of the kernel regression objective function as well. For traditional multi-kernel based algorithm would cost too much time on training model, we develop a very simple and efficient algorithm by adapting feature sets to circulant structure so as to make use of the Fast Fourier Transform (FFT). Thus our algorithm can provide more robust tracking while maintaining real-time effects. To the best of our knowledge, this is the first time the multiple kernel learning algorithms is applied to real-time visual tracking. We evaluate the proposed algorithm on the popular benchmark including 50 image sequences and compare it with 9 state-of-art methods. Implemented in Matlab, the experiment results show that the proposed tracker runs at 45.4 frames per second on an i3 machine and outperforms the state-of-the-art trackers on the benchmark with respect to accuracy. Particularly, the average precision of our algorithm achieves 76.7 % under OPE curve at 20px.

Huan Zhang, Xiankai Chen

Feature Selection Based on Data Clustering

Feature selection is an important step for data mining and machine learning. It can be used to reduce the requirement of data measurement and storage, and defy the curse of dimensionality to improve the prediction performance. In this paper, we propose a feature selection method via mutual information estimation. It avoids the calculation of high-dimensional mutual information by transforming the high-dimensional feature space into one dimension through a novel supervised clustering method. Experimental results on ten benchmark data sets show that: (1) the performances of kNN, naive Bayes classifier, and C4.5 using much less features selected by the proposed method are similar or even better than those on the original data sets with the whole feature set; (2) different from most of state-of-the-art methods which require to setting the number of features to select in prior, the proposed method can automatically determine the proper size of selected feature subsets.

Hongzhi Liu, Zhonghai Wu, Xing Zhang

A Comparison of Local Invariant Feature Description and Its Application

The description of image region draws a lot of attention in the field of computer vision. Recently, many descriptors were proposed for image region description and achieved high achievements. These descriptors are widely used in many fields, such as object recognition, image mosaic, video tracking. In this paper, we first systematically analyze six typical descriptors: SIFT, DAISY, MROGH, MRRID, LIOP and HRI-CSLTP descriptors. Then we conduct experiments in several different situations to evaluate the performance of these descriptors. From the experimental results, we get to make a conclusion and analysis about the advantages and disadvantages of these descriptors. Finally, we make an application of these descriptors in image matching field.

Kaili Shi, Qingwei Gao, Yixiang Lu, Weiguo Zhang, Dong Sun

An Efficient Indexing Scheme Based on K-Plet Representation for Fingerprint Database

Fingerprints are now widely employed in the security fields. A typical police fingerprint database may contain millions of template fingerprints. Consequently, fingerprint indexing plays an essential role to improve the performance of matching such a huge database. In this paper, the efficient index tree based on k-plet local patterns of minutiae for fingerprint database is proposed. The proposed algorithm is of robustness since the k-plet is translation-invariant and rotation-invariant, moreover, the multipath indexing strategy is introduced at the stage of indexing. As well, it is quite fast and effective due to look-up operation instead of complex computation. The performance testing was conducted in the datasets of FVC2002 DB1, NIST DB4 and NIST DB14, which concluded that the proposed algorithm is advantageous for fingerprint indexing since it achieves a high correct index performance with a fairly low penetration rate.

Chaochao Bai, Tong Zhao, Weiqiang Wang, Min Wu

Color Characterization Comparison for Machine Vision-Based Fruit Recognition

In this paper we present a comparison between three color characterizations methods applied for fruit recognition, two of them are selected from two related works and the third is the authors’ proposal; in the three works, color is represented in the RGB space. The related works characterize the colors considering their intensity data; but employing the intensity data of colors in the RGB space may lead to obtain imprecise models of colors, because, in this space, despite two colors with the same chromaticity if they have different intensities then they represent different colors. Hence, we introduce a method to characterize the color of objects by extracting the chromaticity of colors; so, the intensity of colors does not influence significantly the color extraction. The color characterizations of these two methods and our proposal are implemented and tested to extract the color features of different fruit classes. The color features are concatenated with the shape characteristics, obtained using Fourier descriptors, Hu moments and four basic geometric features, to form a feature vector. A feed-forward neural network is employed as classifier; the performance of each method is evaluated using an image database with 12 fruit classes.

Farid García-Lamont, Jair Cervantes, Sergio Ruiz, Asdrúbal López-Chau

Real-Time Human Action Recognition Using DMMs-Based LBP and EOH Features

This paper proposes a new feature extraction scheme for the real-time human action recognition from depth video sequences. First, three Depth Motion Maps (DMMs) are formed from the depth video. Then, on top of these DMMs, the Local Binary Patterns (LBPs) are calculated within overlapping blocks to capture the local texture information, and the Edge Oriented Histograms (EOHs) are computed within non-overlapping blocks to extract dense shape features. Finally, to increase the discriminatory power, the DMMs-based LBP and EOH features are fused in a systematic way to get the so-called

DLE features

. The proposed DLE features are then fed into the

l

2

-regularized Collaborative Representation Classifier (

l

2

-CRC) to learn the model of human action. Experimental results on the publicly available Microsoft Research Action3D dataset demonstrate that the proposed approach achieves the state-of-the-art recognition performance without compromising the processing speed for all the key steps, and thus shows the suitability for real-time implementation.

Mohammad Farhad Bulbul, Yunsheng Jiang, Jinwen Ma

Joint Abnormal Blob Detection and Localization Under Complex Scenes

In this paper, an algorithm is proposed to detect the abnormal event in the form of rectangular blob in global images. Observing the status of the varying blobs, unusual behavior can be monitored and alarmed. A method extracting blobs from crowded video scenes is proposed, the covariance matrix descriptor fuses the image intensity and the optical flow to encode moving information and image characteristics of a blob. After characterizing normal behaviors of blobs or frames in a learning period, the nonlinear one-class SVM algorithm locates the abnormal blobs intra frame. The method is applied to detect abnormal events on several video surveillance datasets, and get promising results.

Tian Wang, Keyu Lai, Ce Li, Hichem Snoussi

Graph Based Kernel k-Means Using Representative Data Points as Initial Centers

The

k

-means algorithm is undoubtedly the most widely used data clustering algorithm due to its relative simplicity. It can only handle data that are linearly separable. A generalization of

k

-means is kernel

k

-means, which can handle data that are not linearly separable. Standard

k

-means and kernel

k

-means have the same disadvantage of being sensitive to the initial placement of the cluster centers. A novel kernel

k

-means algorithm is proposed in the paper. The proposed algorithm uses a graph based kernel matrix and finds

k

data points as initial centers for kernel

k

-means. Since finding the optimal data points as initial centers is an NP-hard problem, this problem is relaxed to obtain

k

representative data points as initial centers. Matching pursuit algorithm for multiple vectors is used to greedily find

k

representative data points. The proposed algorithm is tested on synthetic and real-world datasets and compared with kernel

k

-means algorithms using other initialization techniques. Our empirical study shows encouraging results of the proposed algorithm.

Wuyi Yang, Liguo Tang

Palmprint Recognition Based on Image Sets

In recent years, researchers have found that palmprint is quite a promising biometric identifier as it has the merits of high distinctiveness, robustness, user friendliness, and cost effectiveness. Nearly all the existing palmprint recognition methods are based on one-to-one matching. However, recent studies have corroborated that matching based on image sets can usually lead to a better result. Consequently, in this paper, we present a novel approach for palmprint recognition based on image sets. In our approach, each gallery and query example contains a set of palmprint images captured from a same individual. Competitive code is used for palmprint feature extraction. After the feature extraction process, we use the method of sparse approximated nearest points (SANP) for palmprint image set classification. By calculating the minimum between-set distance, we can set the label of each testing palmprint set as that of the nearest training set. Effectiveness of the proposed approach has been corroborated by the experiments conducted on PolyU palmprint database.

Qingjun Liang, Lin Zhang, Hongyu Li, Jianwei Lu

Face Recognition Using SURF

In recent years, several scale-invariant features have been proposed in literature, this paper analyzes the usage of Speeded Up Robust Features (SURF) as local descriptors, and as we will see, they are not only scale-invariant features, but they also offer the advantage of being computed very efficiently. Furthermore, a fundamental matrix estimation method based on the RANSAC is applied. The proposed approach allows to match faces under partial occlusions, and even if they are not perfectly aligned. Thus based on the above advantages of SURF, we propose to exploit SURF features in face recognition since current approaches are too sensitive to registration errors and usually rely on a very good initial alignment and illumination of the faces to be recognized.

Raúl Cid Carro, Juan-Manuel Ahuactzin Larios, Edmundo Bonilla Huerta, Roberto Morales Caporal, Federico Ramírez Cruz

Research on an Algorithm of Shape Motion Deblurring

According to the feature of fuzzy and the quality reduce background of plate image, this paper presents a Spectrum Estimation Partition Deblur (SEPD) algorithm to restore complex blur image. This algorithm is based on spectrum estimation and extracts the blur kernel information from the dark strips in frequency spectrum, and makes the traditional algorithm better to improve the algorithm’s accuracy and noise capacity. Through the partition recovery strategy, each piece of image is deblurred separately, and then are integrated into a whole image by edge fitting. Through the analysis of actual measurements, the algorithm performs high accuracy and better noise immunity, and also improves the quality of the shape image obviously.

Xia Wu, Hongzhe Xu, Xiaolin Gui, Wen Li, Zhihai Yao

A New Approach for Greenness Identification from Maize Images

Greenness identification from crop growth monitoring images is the first and important step for crop growth status analysis. There are many methods to recognize the green crops from the images, and the visible spectral-index based methods are the most commonly used ones. But these methods can not work properly when dealing with images captured outdoors due to the high variability of illumination and the complex background elements. In this paper, a new approach for greenness identification from maize images is proposed. Firstly, the crop image was converted from RGB color space to HSV color space to obtain the hue and saturation value of each pixel in the image. Secondly, most of the background pixels were removed according to the hue value range of greenness. Then, the green crops were identified from the processed image using the excess green index method and the Otsu method. Finally, all noise objects were removed to get the real crops. The experimental results indicate that the proposed approach can recognized the green plants correctly from the maize images captured outdoors.

Wenzhu Yang, Xiaolan Zhao, Sile Wang, Liping Chen, Xiangyang Chen, Sukui Lu

Image Super-Resolution Reconstruction Based on Sparse Representation and POCS Method

The traditional projection onto convex set (POCS) algorithm can reconstruct a low resolution (LR) image, but it is contradictory in retaining image detail and denoising, so the quality of a reconstructed image is limited. To avoid defects of POCS and obtain higher resolution, the image denoising idea based on sparse representation is led into this paper. Sparse representation can learn well the optimized overcomplete sparse dictionary of an image, which has self-adaptive property to image data and can describe image essential features so as to implement the goal of denoising efficiently. At present, K- singular value decomposition (K-SVD) is the emerging image processing method of sparse representation and has been used widely in image denoising. Therefore, combined the advantages of K-SVD and POCS, a new image ISR method is explored here. In terms of signal noise ratio (SNR) values and the visual effect of reconstructed images, simulation results show that our method proposed has clear improvement in image resolution and can retain image detail well.

Li Shang, Shu-fen Liu, Zhan-li Sun

An Improved Denoising Method Based on Wavelet Transform for Processing Bases Sequence Images

In this article, we present an improved images denoising method for base sequence images. It is based on the multiscale analysis of the images resulting from the à trous wavelet transform decomposition. We define a new thresholding function and use it to improve the denoising performance of the isotropic undecimated wavelet transform (IUWT). The proposed method selects the best suitable wavelet function based on IUWT. The advantages of the new thresholding function are that it is more robust than previous thresholding function, and the convergence of function is more efficient. The experimental results indicate that the proposed method can obtain higher signal-to-noise ratio (SNR) and mean squared error ratio (MSE) than conventional wavelet thresholding denoising methods.

Ke Yan, Jin-Xing Liu, Yong Xu

License Plate Extraction Using Spiking Neural Networks

In this paper, we present an algorithm for license plate detection and extraction using spiking neural networks (SNNs). We propose an SNN for the detection of license plate by simulating the color perception principle in human beings’ visual system, where synchronization of spiking trains are employed as a color detection function and used to detect the license plate according to the difference of color in the license plate’s patch and those in the other image patches. By doing so, we can extract those image regions that are likely to be license plates. And then we use another SNN to produce the edge images of these candidates by simulating the receptive field of orientation in human beings’ visual cortex. Finally, we extract the license plate from these candidates according to the texture difference between a real license plate image and the distracters, where the numbers of strokes in image rows are served as cues for the texture difference. The experimental results show that the proposed biological inspired SNNs are valid in the detection and extraction of license plate.

Qian Du, LiJuan Chen, RongTai Cai, Peng Zhu, TianShui Wu, QingXiang Wu

Pedestrian Detection and Counting Based on Ellipse Fitting and Object Motion Continuity for Video Data Analysis

In order to detect and count pedestrians in different kinds of scenes, this paper put forward a method of solving the problem on video sequences captured from a fixed camera. After preprocessing operations on the original video sequences (Gaussian mixture modeling, three-frame-differencing, image binaryzation, Gaussian filtering, dilation and erosion) we extract the relatively complete pedestrian contours. Then we use the least square ellipse fitting method on those contours that has been extracted, the center of the ellipse is undoubtedly regarded as the tracking point of a pedestrian. With those points, a pedestrian matching pursuit and counting algorithm based on object motion continuity is used for tracking and counting pedestrians, this method can be better used in those scenes which are sparse and rarely obscured. Experiments validate that our pedestrian matching pursuit and counting algorithm has obvious superiorities: good real-time performance and high accuracy.

Yaning Wang, Hong Zhang

Regularized Level Set Method by Incorporating Local Statistical Information and Global Similarity Compatibility for Image Segmentation

This paper presents a regularized level set method for image segmentation, where the local statistical information and global similarity compatibility are both incorporated into the construction of energy functional. By considering the image local statistical information, the proposed model can efficiently segment images with intensity inhomogeneity. To improve the convergence speed, an adaptive stop strategy is proposed. In addition, the distance regularization term is defined with a five power of polynomial function for maintaining the stability during the curve evolution. Finally, experimental results show that our proposed model is efficient for segmenting noisy images, texture images and images with intensity inhomogeneity.

Yu Haiping, Zhang Huali

Sparse Learning for Robust Background Subtraction of Video Sequences

Sparse representation has been applied to background detecting by finding the best candidate with minimal reconstruction error using target templates. However most sparse representation based methods only consider the holistic representation and do not make full use of the sparse coefficients to discriminate between the foreground and the background. Learning overcomplete dictionaries that facilitate a sparse representation of the data as a liner combination of a few atoms from such dictionary leads to state-of-the-art results in image and video restoration and classification. To take these challenges, this paper proposes a new method for robust background detecting via sparse representation. Our method explores both the strength of the well-patch adaptive dictionary learning technique to video frame structure analysis and the robustness background detection by the

l

1

-norm data-fidelity term. By using linear sparse combinations of dictionary atom, the proposed method learns the sparse representations of video frame regions corresponding to candidate particles. The experiments show that the proposed method is able to tolerate the background clutter and video frame deterioration, and improves the existing detecting performance.

Yuhan Luo, Hong Zhang

A New Microcalcification Detection Method in Full Field Digital Mammogram Images

Breast cancer is a great threat for women around the world. Mammography is the main approach for early detection and diagnosis. Microcalcification (MC) in mammograms is one of the important early signs of breast cancer. Their accurate detection is important in computer-aided detection (CADe). In this paper, we proposed a new Microcalcification detection method for full field digital mammograms (FFDM) by integrating Possibilistic Fuzzy c-Means (PFCM) clustering algorithm and weighted support vector machine (WSVM). The method includes a training process and a testing process. In the training process, possible microcalcification regions are located and extracted. Extracted features are selected with mutual information based technique. Positive and negative samples are weighted with PFCM and used to train a weighted SVM. A similar procedure is performed on test images. The proposed method is evaluated on a database of 410 clinical mammograms and compared with a standard unweighted support vector machine classifier.

Xiaoming Liu, Ming Mei, Weiwei Sun, Jun Liu

Forensic Detection of Median Filtering in Digital Images Using the Coefficient-Pair Histogram of DCT Value and LBP Pattern

Looking for modification traces of digital media is of great value for forensic analysis. The median filter can be used to remove the fingerprints left by other image operations, and the detection of median filtering has become more and more significant. In this paper, a new detector for median-filtering operation is proposed. In the method, the image features combined by LBP (Local Binary Pattern) and coefficient-pair histogram in DCT (Discrete Cosine Transform) domain are firstly extracted; then classifier SVM is used to train the authentic and median-filtered image; lastly, some suspicious images are used to test the effectiveness of the proposed scheme. Large amounts of experiments show that the proposed method can detect median filtering under a variety of scenarios, and further more it has letter robustness against JPEG post-compressed image, this outperforms the existing state-of-the-art method.

Yun-Ni Lai, Tie-Gang Gao, Jia-Xin Li, Guo-Rui Sheng

An Image Enhancement Method Based on Edge Preserving Random Walk Filter

Some previous edge preserving smoothing methods suffer from halo artifacts when they are applied for image enhancement. In this paper, an edge preserving random walk filter is proposed, our method suffers free from artifacts. Unlike previous methods, the proposed method is able to obtain a smoothing result by just solving a system of linear equation. The proposed filter is then adopted to design an image enhancement algorithm. By just amplifying and adding the detail layer to the base layer, the algorithm can produce a satisfactory result. The simulation results demonstrate that our approach performs much better than other existing techniques.

Zhaobin Wang, Hao Wang, Xiaoguang Sun, Xu Zheng

Latent Fingerprint Segmentation Based on Sparse Representation

Latent fingerprints are the finger skin impressions which are left at the scene of a crime by accident. They are usually of poor quality with weak fingerprint ridge flows and various overlapping irrelevant patterns. It is still a challenging problem for automatic latent fingerprint processing and recognition. Latent fingerprint segmentation, which segments the fingerprint ridge area from complex backgrounds, is an important preprocessing step for latent fingerprint recognition. This paper proposes a latent fingerprint segmentation algorithm based on sparse representation. First, the total variation (TV) model is used to decompose a latent image into two components: texture and cartoon. The texture component, which contains the weak fingerprint ridge and valley structures, is used for further processing, while the cartoon component mainly consisting of the irrelevant information is discarded as noises. Then, we compute the sparse representation of the texture image against the dictionary constructed by a set of Gabor elementary functions. Since the sparse coefficients measure the weights of the basis atoms in fingerprint representation, an image quality measure is computed from the sparse coefficients, which evaluate how well the texture image can be sparsely reconstructed from the basis atoms. Finally, this image quality measure is used for fingerprint segmentation. We test the proposed method on the NIST SD27 latent fingerprint database. Experimental results and comparisons demonstrate the effectiveness of the proposed method.

Kaifeng Wei, Xiaoying Chen, Manhua Liu

Pose Estimation for Vehicles Based on Binocular Stereo Vision in Urban Traffic

Extensive research has been carried out in the field of driver assistance systems in order to increase road safety and comfort. We propose a pose estimation algorithm based on binocular stereo vision for calculating the pose of on-road vehicles and providing reference for the decision of driving assistant system, which is useful for behavior prediction for vehicles and collision avoidance. Our algorithm is divided into three major stages. In the first part, the vehicle is detected and roughly located on the disparity map. In the second part, feature points on the vehicle are extracted by means of license plate detection algorithm. Finally, pose information including distance, direction and its variation is estimated. Experimental results prove the feasibility of the algorithm in complex traffic scenarios.

Pengyu Liu, Fei Wang, Yicong He, Hang Dong, Haiwei Yang, Yang Yang

Robust Segmentation of Vehicles Under Illumination Variations and Camera Movement

Vision-based vehicle detection and segmentation in intelligent transportation systems, particularly under outdoor illuminations, camera vibration, cast shadows and vehicle variations is still an area of active research for analysis and processing of traffic data. This paper proposes an effective scheme that improves Gaussian mixture model (GMM) for non-stationary temporal distributions through dynamically updating the learning rate at each pixel. In this proposed technique, sleeping foreground pixels and slow moving vehicles cannot become the part of background model that also does not lead to extra computational cost as compare to other methods that are proposed in the literature. Sudden illumination change is also captured in this technique. Vision based system cannot be efficient without fixing of camera vibration, so movement of camera is adjusted based on clues from background model. At the end, shadows are removed from detected vehicles through applying a new recursive method in dark regions. Experimental results demonstrate the robustness and high level performance of the proposed adaptive foreground extraction algorithm under illumination variations compared to state-of-the-art methods.

Zubair Iftikhar, Prashan Premaratne, Peter Vial, Shuai Yang

Binarization Chinese Rubbing Images Using Gaussian Mixture Model

Rubbings are important components of ancient Chinese books, and are the main source for people to learn, study, and research history. Image segmentation plays a crucial role in extracting useful information and characteristics of Chinese character from the rubbing images. In this paper, binarization using a Gaussian Mixture Model (GMM) with 2 components for representation of background and foreground distribution in a Chinese rubbing image has been proposed. To model the likelihood of each pixel belonging to foreground or background, a foreground and background color model are learned from three color bands samples that using RGB color space. The standard Expectation-Maximisation (EM) algorithm had been used to estimate the GMM parameters. Experimental results on real rubbing images validate the effectiveness of the model when working with Chinese rubbing images.

Zhi-Kai Huang, Fang Wang, Jun-Mei Xi, Han Huang

Locally Linear Representation Manifolds Margin

In this paper, a novel supervised multiple manifolds learning method is presented for dimensionality reduction, which is titled locally linear representation manifold margin (LLRMM). In the proposed LLRMM, both an inter-manifold graph and intra-manifold graph are constructed, where any point in the inter-manifold graph must select neighbors from other manifolds while the neighborhood in the intra-manifold graph are composed of samples from the same manifold. Then the least locally linear representation technique is introduced to optimize the reconstruction weights as well as the corresponding inter-manifold scatter and intra-manifold scatter, based on which manifolds margin can be reasoned. At last, a discriminant subspace is explored. Experiments on some benchmark face data sets have been conducted and experimental results show that the proposed method outperforms some related state-of-the-art dimensionality reduction methods.

Bo Li, Yun-Qing Wang, Lei Lei, Zhang-Tao Fan

Modified Sparse Representation Based Image Super-Resolution Reconstruction

A modified sparse representation based image super-resolution reconstruction (ISR) is discussed in this paper. The edge features of high resolution (HR) image patches and the gradient and texture features of low resolution (LR) image patches are considered in our method. Meanwhile, features of LR image patches are classified by extreme learning machine (ELM) classifier. Further, For image patches’ features classified, the fast sparse coding (FSC) algorithm based K-SVD sparse representation is used to train sparse dictionaries. And utilized these dictionaries, LR images can be super-resolution reconstructed well. Simulation results show that our method has clear improvement in visual effect and retain well image detail.

Li Shang, Pin-gang Su, Zhan-li Sun

Planning Feasible and Smooth Paths for Simulating Realistic Crowd

A very important challenge in many virtual applications is to plan feasible, smooth and congestion-free paths for virtual agents in dynamic and complex environments. The agents should move towards their destinations successfully while avoiding the collisions with other agents and static and dynamic obstacles. In this paper, we propose a novel approach for realistic path planning. We first create a navigation mesh for the walkable regions in a two-dimensional environment. Then an A* search on this graph determines a series of connected meshes for agents to go through from the start position to the goal position and furthermore, the walkable corridor whose radii equal to maximum clearance to the obstacles is built based on backbone path derived from the inflection point method and Catmull-Rom spline. Finally, a local collision avoidance algorithm is integrated to guarantee that agents navigating in the corridor do not collide with other agents and dynamic obstacles. Our experiments show that we can compute feasible, smooth and realistic paths for agents situated in dynamic environments in real time.

Libo Sun, Lu Ding, Wenhu Qin

A Comparative Investigation of PSG Signal Patterns to Classify Sleep Disorders Using Machine Learning Techniques

Patients with Non-Communicable Diseases (NCDs) are increasing around the globe. Possible causes of the NCDs are continuously being investigated. One of them is a sleep disorder. In order to detect specific sleep disorders, the Polysomnography (PSG), is necessary. However, due to the lack of the PSG in many hospitals, researchers attempt to discover alternative approaches. This article demonstrates comparisons of sleep disorder classifications using machine learning techniques. Three main machine learning techniques have been compared including Classification And Regression Tree (CART),

k

-Mean Clustering (KMC) and Support Vector Machine (SVM). The SVM achieves the best classification results in NREM-1 and NREM-2. The CART performs superior in NREM-3 and REM. Implications in terms of medical diagnosis, there are two main selected features, SaO2 and Pulse, based on the CART in all of the sleep stages. The features may be pieces of evidences to predict various types of sleep disorders.

Thakerng Wongsirichot, Anantaporn Hanskunatai

A Multi-valued Coarse Graining of Lempel-Ziv Complexity and SVM in ECG Signal Analysis

Lempel-Ziv (LZ) complexity method has been widely applied to detection ventricular tachycardia (VT) and ventricular fibrillation (VF). The coarse-graining process (Quantization levels,

$$ L $$

) plays an important role in the LZ complexity measure analysis. In this paper, we present a multi-valued coarse-graining process approaches (

$$ L > 2 $$

), our test shows that this algorithm is superior to the two-valued coarse-graining of LZ complexity approaches (

$$ L = 2 $$

) in VT and VF separation. Furthermore, we used support vector machine (SVM) classifier to discriminate VF and VT. Using the complexity as a feature to input classifiers can significantly improve the classification results. Particularly, optimum performance is achieved at a 4-second length.

Deling Xia, Qingfang Meng, Yuehui Chen, Zaiguo Zhang

A Survey of Multiple Sequence Alignment Techniques

Multiple sequence alignment (MSA) is a basic step in many bioinformatics analyses, and also a NP-hard problem. In order to improve the speed, accuracy and cater to the requirement of large-scale sequences alignment, a wide variety of MSA methods and softwares have been subsequently developed. In this article, we will systematically review the wildly used methods and introduce their practical results on the benchmark Balibase 3.0 references. We come to the conclusion that computational complexity still is the bottleneck of MSA. We also consider future development of MSA methods with respect to applying of more different technologies and the prospect of parallelization of MSA.

Xiao-Dan Wang, Jin-Xing Liu, Yong Xu, Jian Zhang

Analyzing the Genomes of Coxsackievirus A16 and Enterovirus 71 in Relation to Hand, Foot and Mouth Disease(HFMD) Using Apriori Algorithm, Decision Tree and Support Vector Machine (SVM)

Hand, foot and mouth disease (HFMD), caused by highly infectious intestinal viruses of either coxsackievirus A16 (CVA16) or enterovirus 71 (EV71), is a common children syndrome featured by mild fever, spots and bumps that blister the skin of hands, oral cavity in mouth, feet and sometimes to the extent of buttocks and genitalia. Though CVA16 and EV71 both cause the HFMD, the intensity of each symptom is obviously different. Normal cases are HFMD by CVA16, which are typically characterized by mild symptoms usually treated in 6−8 days. Conversely, HFMD by EV71 results severe neural disorders and various influenza complications that even cause death. Currently, no vaccine is available to protect individuals from infection by the viruses that cause HFMD. In order to investigate why these two viruses have too much differences in degree of symptoms and compare the inner relationships to the medical effects, we analyzed the genomes of CVA16 and EV71 by using apriori algorithm, decision tree and support vector machine (SVM). Therefore, by comparing the genomes of each virus, we found out better results for analyzing the relationship between the two and state the potential of developing medical remedy in DNA point of view.

Chaeyun Jung, Yonghyun Park, Seunghui Han, Taeseon Yoon

A Utility Function Based Resource Allocation Method for LEO Satellite Constellation System

Low Earth Orbit (LEO) satellite constellation system has been regarded as a very promising satellite mobile communication system for its low propagation delay, global coverage for communication and mature technologies. However, considering its crucial but limited resources on satellites, efficient resource allocation methods are needed to guarantee the network carrying capability under specific requirements. In this paper, we put forward a utility function based resource allocation method for LEO satellite constellation system. We first utilize utility function to represent the resource acquisition satisfaction of adjacent satellites. The bigger the utility is the higher satisfaction the adjacent satellites can get. In addition, we adopt the improved Multiple Population Cloud Differential Evolution Algorithm (MPCDEA) to solve the resource allocation problem, which belongs to the nonlinear mixed integer programming problem. Finally, we evaluate the proposed utility function based resource allocation method according to the topologies of Iridium and Globalstar systems and quantify it with performance indexes like throughput capacity and network capacity. Evaluation results show that our method is feasible and effective.

Fangfang Yuan, Xingwei Wang, Fuliang Li, Min Huang

A Novel Naive Bayes Classifier Model Based on Differential Evolution

Naive Bayes (NB) classifier is a simple and efficient classifier, but the independent assumption of its attribute limits the application of the actual data. This paper presents an approach called Differential Evolution-Naive Bayes (DE-NB) which takes advantage of combining differential evolution with naive Bayes for attribute selection to improve naive Bayes classifier. This method applies DE firstly to search out an optimal subset of attributes reduction in the original attribute space, and then constructs a naive Bayes classifier on the gotten subset of the attributes reduction. Nineteen experimental results on UCI datasets distinctly show that compared with Cfs-BestFirst algorithm, NB algorithm, Support Vector Machine (SVM) algorithm, Decision Tree (C4.5) algorithm, K-neighbor (KNN) algorithm, the proposed algorithm has higher classification accuracy.

Jun Li, Guokang Fang, Bo Li, Chong Wang

Chaotic Iteration Particle Swarm Optimization Algorithm Based on Economic Load Dispatch

To solve the non-convex and non-linear economic dispatch problem efficiently, a chaotic iteration particle swarm optimization algorithm is presented. In the global research of particle swarm optimization and local optimum, ergodicity of chaos can effectively restrain premature. To balance the exploration and exploitation abilities and avoid being trapped into local optimal, a new index, called iteration best, is incorporated into particle swarm optimization, and chaotic mutation with a new Tent map imported can make local search within the prior knowledge, a new strategy is proposed in iteration strategy. The algorithm is validated for two test systems consisting of 6 and 15 generators. Compared with other methods in this literature, the experimental result demonstrates the high convergency and effectiveness of proposed algorithm.

Zhenghong Yu, Fengli Zhou

Training Artificial Neural Network Using Hybrid Optimization Algorithm for Rainfall-Runoff Forecasting

In this paper, a hybrid optimization algorithm is proposed to train the initial connection weights and thresholds of artificial neural network (ANN) by incorporating Simulated Annealing algorithm (SA) into Genetic Algorithm (GA), and then the Back Propagation (BP) algorithm is applied to adjust the final weights and biases, namely HGASA-ANN. Finally, a numerical example of daily rainfall-runoff data is used to elucidate the forecasting performance of the proposed HGASA-ANN model. The GASA is employed to accelerate the training speed and helps to avoid premature convergence and permutation problems. The HGASA-NN can make use of not only strong global searching ability of the GASA, but also strong local searching ability of the BP algorithm. The forecasting results indicate that the proposed model yields more accurate forecasting results than the back-propagation neural network and pure GA training artificial neural network. Therefore, the HGASA-ANN model is a promising alternative for rainfall-runoff forecasting.

Jiansheng Wu, Chengdong Wei

A Novel Swarm Intelligence Algorithm Based on Cuckoo Search Algorithm (NSICS)

Cuckoo Search algorithm (CS) is swarm intelligence based algorithm motivated by nature. This algorithm is based on brood parasitism of some cuckoo species and has high capability of global search. Therefore, the global optimum can be figured out with higher probability. This paper proposes a novel meta-heuristic approach, called NSICS, based on CS. NSICS is able to explore not only the search space on global scale but also around the optimum on local scale more efficiently. Consequently, more accurate results can be obtained. To approach these purposes, three operators of Eggs laying, lévy fights and Move are applied. Experiments are studied on thirteen common benchmark functions among unimodal, multimodal, shifted and shifted rotated classes and then compared with CS, GPSO, SFLA and GSA algorithms. These algorithms are chosen from swarm intelligence based, bio-inspired based and chemistry and physics based algorithms’ category. The simulations indicate the proposed algorithm has satisfactory performance.

Nazanin Fouladgar, Shahriar Lotfi

Robust PCA-Based Genetic Algorithm for Solving CNOP

Conditional nonlinear optimal perturbation (CNOP) has been widely used in the predictability and sensitivity studies of the weather or climate models. The popular solution to the CNOP is the adjoint-based method. However, many numerical models have no adjoint models, thus bringing about a limitation to the CNOP applications. To avoid the adjoint models, we propose the robust PCA-based genetic algorithm for solving the CNOP (RGA_CNOP). To demonstrate the validity of the proposed method, it is applied to the CNOP of the Zebiak-Cane (ZC) model, and compared with the adjoint-based method. Experimental results show the RGA_CNOP can obtain approximate results to the adjoint-based method.

Shicheng Wen, Shijin Yuan, Bin Mu, Hongyu Li

Exploring Three Emotion Induction Procedures and Their Effect on E-learners’ Language Learning

This study explores the impact of positive, negative and neutral emotion induction procedures on Chinese adult e-learners’ language learning. Thirty students from each of the three groups were selected as the subjects, with each group receiving one of the three treatments. The subjects attended some online lecturing sessions on English tenses and were assigned a pretest and a posttest. Data were also collected through simultaneous recording of the lecturing. The results of the posttest show that the positive treatment can generate the most facilitating impact on learning, the neutral one can also facilitate learning but not as much as the positive one, and the negative one may hinder learning.

Zigang Ge

Approximate Bit-Vector Algorithms for Hashing-Based Similarity Searches

Similarity search, or finding approximate nearest neighbors, is becoming an increasingly important tool to find the closest matches for a given query object in large scale database. Recently, learning hashing-based methods have attracted considerable attention due to their computational and memory efficiency. The basic idea of these approaches is to generate binary codes for data points which can preserve the similarity between any two of them. In this paper, we propose a novel algorithm named Approximate Bit-Vector (ABV) for hashing-based similarity search. ABV algorithm map data points into Hamming space and integrate with hash functions for fast similarity or

k

-NN search. Extensive experimental results over real large-scale datasets demonstrate the superiority of the proposed approach.

Ling Wang, Tie Hua Zhou, Zhen Hong Liu, Zhao Yang Qu, Keun Ho Ryu

Dynamic Hand Gesture Recognition Using Centroid Tracking

In many dynamic hand gesture recognition contexts, time information is not adequately used. The extracted features of dynamic gestures usually do not carry explicit information about time in gesture classification. This results in under-utilized data for more important accurate classification. Another disadvantage is that the gesture classification is then confined to only simple gestures. We have overcome these limitations by introducing centroid tracking of hand gestures that captures and retains the time sequence information for feature extraction. This simplifies the classification of dynamic gestures as movement in time helps efficient classification without burdensome processing.

Prashan Premaratne, Shuai Yang, Peter Vial, Zubair Ifthikar

Conditional Matching Preclusion Sets for an Mixed-Graph of the Star Graph and the Bubble-Sort Graph

The conditional matching preclusion number of a graph is the minimum number of edges, whose deletion results in a graph with no isolated vertices that has neither perfect matchings nor almost-perfect matchings. Any such optimal set is called an optimally conditional matching preclusion set. The conditional matching preclusion number is one of the parameters to measure the robustness of interconnection networks in the event of edge failure. The star graph and the bubble-sort graph are one of the attractive underlying topologies in a multiprocessor system. In this paper, we investigate a class of Cayley graphs which are combined with the star graph and the bubble-sort graph, and give all the optimally conditional matching preclusion sets for this class of graphs.

Yunxia Ren, Shiying Wang

A Method to Select Next Hop Node for Improving Energy Efficiency in LEAP-Based WSNs

In wireless sensor networks, sensors have stringent energy and computation requirements as they must function unattended. The sensor nodes can be compromised by adversaries who attack network layers such as in sinkhole attacks. Sinkhole attacks have the goal of changing routing paths and snatching data surrounding the compromised node. A localized encryption and authentication protocol (LEAP) observes different types of messages exchanged between sensors that have different security requirements to cope with the attack. Even though this original method excels in security communication using multiple keys, the data is transmitted without optimal selection of the next nodes. In this paper, our proposed method selects the optimal next node based on a fuzzy logic system. We evaluated the energy and security performances of our method against sinkhole attack. Our focus is to improve energy efficiency and maintain the same security level as compared to LEAP. Experimental results indicated that the proposed method saves up to 5 % of the energy while maintaining the security level against the attack as compared to LEAP.

Su Man Nam, Tae Ho Cho

An Efficient Topology-Based Algorithm for Transient Analysis of Power Grid

In the design flow of integrated circuits, chip-level verification is an important step that sanity checks the performance is as expected. Power grid verification is one of the most expensive and time-consuming steps of chip-level verification, due to its extremely large size. Efficient power grid analysis technology is highly demanded as it saves computing resources and enables faster iteration. In this paper, a topology-base power grid transient analysis algorithm is proposed. Nodal analysis is adopted to analyze the topology which is mathematically equivalent to iteratively solving a positive semi-definite linear equation. The convergence of the method is proved.

Lan Yang, Jingbin Wang, Lorenzo Azevedo, Jim Jing-Yan Wang

Implementation of Leaf Image Recognition System Based on LBP and B/S Framework

Plant identification system is on the basis of the previous, through continuous optimizing all aspects of the algorithm to improve efficiency and accuracy of the algorithm. For feature extraction, since the local binary pattern was proposed in the past decades, it has been widely used in computer vision to describe the feature for image classification such as image recognition, motion detection and medical image analysis. According to accuracy of the descriptor always fluctuates with different samples, some improved pattern of LBP has been presented in papers. Complete Local Binary Pattern (CLBP) is an optimized version which set an additional magnitude value to local differences. This paper shows extensive experiments of implement the LBP derivatives for plants texture identification. Finally realize an online system to identify what kind of the plant image user uploaded based on LBP descriptor.

Sen Zhao, Xiao-Ping Zhang, Li Shang, Zhi-Kai Huang, Hao-Dong Zhu, Yong Gan

Using Additive Expression Programming for System Identification

The system identification is crucially important process, which could develop the mathematical representation of physical system from observed data. In this paper, a new model, called additive expression tree (AET) model is proposed to encode the linear and nonlinear systems. A new structure-based evolutionary algorithm and artificial bee colony (ABC) are used to optimize the architecture and parameters of additive expression tree model, respectively. Experimental results demonstrate that our proposed model and hybrid approach could identify the linear/nonlinear systems effectively.

Bin Yang

Supervised Feature Extraction of Hyperspectral Image by Preserving Spatial-Spectral and Local Topology

Manifold learning, as a promising tool for nonlinear dimensionality reduction of hyperspectral image (HSI) data, has drawn great research interests in the remote sensing community. It can extract meaningful and low-dimensional features underlying complex HSI data, which is useful in classification of ground targets. However, there are two limitations with current approaches, few considerations of spatial information and lack of explicit mapping relationship. In this paper, we propose a supervised spatial-spectral local topology preserving embedding (sssLTPE) method for efficient feature extraction of HSI, which owns two merits. First, spatial and spectral information at each pixel is integrated by an intuitive strategy. Second, an explicit and nonlinear mapping relationship is provided to effectively map unlabeled data to learned feature space. Experiments conducted on benchmark data set demonstrate that high classification accuracy can be obtained by using the features extracted by sssLTPE.

Peng Zhang, Haixia He, Zhou Sun, Chunbo Fan

A New Learning Automata Algorithm for Selection of Optimal Subset

A new class of learning automata for the purpose of learning the optimal subset of actions has been proposed to fulfill the demand of application such as allocation and global optimization. Learning automata are capable of dealing with multiple choice problems if some modifications are made on current algorithms. This paper discusses on how to adapt current LA algorithms to the new purpose and introduces a new kind of learning automata. The proposed automata take advantage of LELA, whose original updating schemes favor the purpose of selecting multiple actions and thus acquire faster rate of convergence than the existing automata for selecting optimal subset of actions to the best of our knowledge. Additionally, extensive simulation results are presented to compare the performance between the proposed algorithm and the existing ones. The results show that the proposed automata outperform the other automata.

Xinyi Guo, Wen Jiang, Hao Ge, Shenghong Li

The Recent Developments and Comparative Analysis of Neural Network and Evolutionary Algorithms for Solving Symbolic Regression

Symbolic regression (SR) is one of the research fields in data mining, how to use scientific and appropriate methods to study SR is a difficult problem. The traditional methods used in SR mainly focus on the models such as genetic programming (GP), the article applies the gene expression programming (GEP) and neural network (NN) to this field, in order to correctly compare the advantages and disadvantages of the three methods, some relevant works have been done. This paper first briefly introduces the NN and evolutionary algorithms including GP and GEP, their design steps and recent developments, and applies these algorithms to SR, then uses the algorithms to solve SR and makes comparison analysis, and draws some conclusions in the experiment condition: the performance of NN and evolutionary algorithms change dramatically for solving this problem; GP and GEP fluctuate greatly compared with NN, the used time is also less, and NN shows better stability and result.

Xueshi Dong, Wenyong Dong, Yunfei Yi, Yajie Wang, Xiaosong Xu

Inferring Large Gene Networks with a Hybrid Fuzzy Clustering Method

To tackle the scalability problem in reverse engineering gene networks, this study presents an approach with two phases: gene clustering and network reconstruction. For gene clustering, a hybrid data and knowledge-driven method is developed to calculate similarity between genes. In the network reconstruction procedure, a Boolean network model is inferred from gene clusters. A series of experiments are conducted to investigate the effect of the hybrid similarity measure in gene clustering and network reconstruction. The results prove the feasibility and effectiveness of the proposed approach.

Chung-Hsun Lin, Yu-Ting Hsiao, Wei-Po Lee

A Novel Algorithm for Classifying Protein Structure Familiar by Using the Graph Mining Approach

Protein structural classification is critical in bioinformatics. In this study, a simple and connected graph was used to represent a 3D protein structure in which each node represented an amino acid and each edge represented a contact distance between two amino acids. The B-factor (atomic displacement parameters) was then used to substantially reduce the number of nodes and edges in each graph representation. A graph mining approach was applied to determine the critical subgraphs among these graphs, which can be applied to classify protein structural families. An experimental study was conducted in which characteristic substructural patterns were identified in several protein families in the SCOP database.

Sun-Yuan Hsieh, Chia-Wei Lee, Zong-Ying Yang, Heng-Wei Wang, Jun-Han Yu

Predicting Helix Boundaries of α-Helix Transmembrane Protein with Feedback Conditional Random Fields

Transmembrane proteins play an important role in cellular energy production, signal transmission, metabolism. Existing machine learning methods are difficult to model the global correlation of the membrane protein sequence, and they also can not improve the quality of the model from sophisticated sequence features. To address these problems, in this paper we proposed a novel method by a feedback conditional random fields (FCRF) to predict helix boundaries of

α

-helix transmembrane protein. A feedback mechanism was introduced into multi-level conditional random fields. The results of lower level model were used to calculate new feedback features to enhance the ability of basic conditional random fields. One wide-used dataset DB1 was used to validate the performance of the method. The method achieved 95 % on helix location accuracy. Compared with the other predictors, FCRF ranks first on the accuracy of helix location.

Kun Wang, Hongjie Wu, Weizhong Lu, Baochuan Fu, Qiang Lü, Xu Huang

Prediction of Protein Structure Classes

Prediction of protein structural are crucial in Bioinformatics. More and more evidences demonstrate that an great number of prediction methods has been employed to predict these structures based on the sequences of protein and biostatistics. The accuracy of such methods, nevertheless, is strongly affected by the efficiency and the robustness of classification model and other several factors. In our present research, the features based on the correlation coefficient of dipeptide or polypeptide were put forward. For one thing, flexible neutral tree(FNT), a novel classification model which is a variable structure neural network, is employed as the base classifiers. For another, the alterable tree structure based on FNT, such model may take advantage of the selection of available information, which aimed at the improvement of efficiency. It is important to find out the tree structural of protein structure classification model. To examine the performance of such method, ASTRAL, 1189 and 640 are selected as benchmark datasets of protein tertiary structure. Fortunately, the results show that a higher prediction accuracy compared with other methods. With the selected features running in the flexible neutral tree, several redundant information of features may be cut off and the accuracy of such model may be improved in some degree and the time of running such model could be hold down.

Wenzheng Bao, Dong Wang, Fanliang Kong, Ruizhi Han, Yuehui Chen

An Integrated Computational Schema for Analysis, Prediction and Visualization of piRNA Sequences

PIWI-interacting RNAs (piRNAs) are endogenously originated predominantly germline oriented newly described class of small non-coding RNAs or transcripts. piRNAs are found to be crucial ones in rendering translational arrest of proteins thereby protecting the genome germline integrity from invasive transposable elements. piRNAs are demanding more attention because of its potential role in the process of spermatogenesis and male infertility. Though there exist several computational approaches to predict piRNA sequences, a parameter based piRNA prediction strategy was not attempted yet. Understanding this scenario, a comprehensive computational schema has been developed based on Bayes and Tree classifiers. The proposed method provides an integrated platform to analyze, predict and visualize piRNA dataset from other noncoding RNAs in a multi-threaded environment. Moreover, a comparative study of different classification algorithms applicable to piRNA predictions is presented here.

Anusha Abdul Rahiman, Jithin Ajitha, Vinod Chandra

Backmatter

Weitere Informationen

Premium Partner

    Bildnachweise