Emerging Intelligent Computing Technology and Applications

The objects of traditional plant identification were too broad and the classification features of it were usually not synthetic and the recognition rate was always slightly low. This paper gives one recognition approach based on supervised locally linear embedding (LLE) and K-nearest neighbors. The recognition results for thirty kinds of broad-leaved trees were realized and the average correct recognition rate reached 98.3%. Comparison with other recognition method demonstrated the proposed method is effective in advancing the recognition rate.

This paper presents a decision support based, dynamic approach to optimal threat evaluation and defensive resource scheduling. The algorithm provides flexibility and optimality by swapping between two objective functions, based on preferential and subtractive defense strategies, as and when required. Analysis part of this paper presents the strengths and weaknesses of the proposed algorithm over an alternative greedy algorithm as applied to different offline scenarios.

In this paper, how to construct quantum quasi-cyclic (QC) low-density parity-check (LDPC) codes is proposed. Using the proposed approach, some new quantum codes with various lengths and rates of no cycles-length 4 in their Tanner graph are designed. In addition, the presented quantum codes can be efficiently constructed with large codeword length. Finally, we show the decoding of the proposed quantum QC LDPC.

Using unsupervised algorithms to cluster for diagnosis information data is a mainstream and difficult area of TCM clinical research, and the optimal symptoms’ number of the syndrome is even more difficult to gain. However, there is no relevant and effective research on it yet. An unsupervised clustering algorithm is proposed based on the concepts of complex system entropy and contribution degree in this work. The algorithm is based on the familiar unsupervised complex system entropy cluster algorithm, simultaneously, it introduces contribution degree to self-adaptively select the symptoms’ number. This work carried out three clinical epidemiology surveys about depression, chronic renal failure and chronic hepatitis b, and obtained 1787 cases, each of which has measurements for 76 symptoms. The algorithm discovers 9 patterns, and 6 of them fit the syndrome in clinic. Therefore, we conclude that the algorithm provides an effective solution to discover syndrome from symptoms.

There are always two main problems in identification of human beings through their ear images: 1- If distances of the individual from camera changes, the sizes of ears in the saved images are varied in proportion to this distance. 2- If head of people in taken images is tilted upwards or downwards, this causes ear images of these people rotate in proportion to saved ear images in database. In both of these cases, all identification systems do not work properly. In this article, we proposed a new method for normalizing human ear images by detecting the rotation and scaling variation, and normalizing the ear images accordingly. Our proposed method works well on all ear databases and all ear images (either left or right) which have been taken from front side of the ears. Our method provides high performance to the biometric identification systems to identify human being, even when the images of human ears are taken from long distance with small scale.

In order to deal with the interaction between genes effectively, a kernel technology was adopted into a subspace method in our study. A linear subspace classifier was generalized to a nonlinear kernel subspace classifier by using a kernel principle component analysis method to constitute nonlinear feature subspaces. Because DNA microarray data have characteristics of high dimension, few samples and strong nonlinearity, three types of classifiers based on kernel machine learning method were designed, i.e., support vector machine (SVM), kernel subspace classifier (KSUB-C) and kernel partial least-squares discriminant analysis (KPLS-DA). But the performances of these classifiers lie on the optimum setting of kernel functions and parameters. Therefore, to avoid the difficulty of selecting optimal kernel functions and parameters and to further improve the accuracy and generalization property of the classifiers, an ensemble classifier based on kernel method for multi-situation DNA microarray data was proposed by adopting the idea of ensemble learning. The ensemble classifier combines the classification results of the SVM, KSUB-C and KPLS-DA classifiers. Experimental results involving three public DNA microarray datasets indicate that the proposed ensemble classifier has high classification accuracy and perfect generalization property.

Signaling pathway construction is one of hotspots in the present bioinformatics. A novel approach where priori knowledge is fused is proposed, called Dk-NICO, where partial missing regulation relationships and regulation directions are used as data samples, and biological experiment result as priori knowledge, while HMM is used as a model for reconstructing the signaling pathway, so as to predict signaling pathway. By reconstructing MAPK pathway, it is showed that the proposed approach not only is capable of predicting gene regulation relationships, but also is capable of identifying gene regulation directions. Moreover, we apply the approach to MAPK pathway reconstruction in the case of no priori knowledge and demonstrate that, by introducing priori knowledge from direct biochemical reaction experiment, the prediction accuracy is improved.

Recent improvements in high-throughput proteomics technology have produced a large amount of time-series gene expression data. The data provide a good resource to uncover causal gene-gene or gene-phenotype relationships and to characterize the dynamic properties of the underlying molecular networks for various biological processes. Several methods have been developed for identifying the molecular mechanisms of regulation of genes from the data, but many of the methods consider static gene expression profiles only. This paper presents a new method for identifying gene regulations from the time-series gene expression data and for visualizing the gene regulations as dynamic gene regulatory networks. The method has been implemented as a program called DRN Builder (Dynamic Regulatory Network Builder; http://wilab.inha.ac.kr/drnbuilder/) and successfully tested on actual gene expression profiles. DRN Builder will be useful for generating potential gene regulatory networks from a large amount of time-series gene expression data and for analyzing the identified networks.

In this paper, a novel signature verification algorithm based on spatio-temporal correlation and an improved template selection method are proposed. The proposed algorithm computes the spatio-temporal matrices of signatures and considers them as images, then applies image processing methods to obtain features. The spatio-temporal matrices will enlarge the variations of the signatures, so an improved template selection method considering the intra-class variation of enrollment signatures is also adopted. The algorithms are validated on the SVC 2004 database and inspiring results are obtained.

According to Barrow and Tenenbaum’s theory, an image can be decomposed into two images: a reflectance image and an illumination image. This midlevel description of images attracts more and more attentions recently owing to its application in computer vision, i.e. facial image processing and face recognition. However, due to its ill-posed characteristics, this decomposition remains difficult. In this paper, we concentrate on a slightly easier problem: given a simple frontal facial image and a learned near infrared image, could we recover its reflectance image? Experiments show that it is feasible and promising. Based on extensive study on hyperspectral images, skin color model and Quotient Image, we proposed a method to derive reflectance images through division operations. That is to divide visual frontal face images by learned near infrared images which are generated by super-resolution in tensor space. With the operation on grey distribution of frontal facial images, the results after division can represent the reflectance of skin, rarely bearing any illumination information. Experimental results show that our method is reasonable and promising in image synthesis, processing and face recognition.

In this paper, a new method for gender recognition via gait silhouettes is proposed. In the feature extraction process, Radon transform on all the 180 angle degrees is applied to every silhouette to construct gait templates and the initial phase of each silhouette in an entire gait cycle is also associated to the templates representing dynamic information of walking. Then the Relevant Component Analysis (RCA) algorithm is employed on the radon-transformed templates to get a maximum likelihood estimation of the within class covariance matrix. At last, the Mahalanobis distances are calculated to measure gender dissimilarity in recognition. The Nearest Neighbor (NN) classifier is adopted to determine whether a sample in the Probe Set is male or female. Experimental results in comparison to state-of-the-art methods show considerable improvement in recognition performance of our proposed algorithm.

It is of great importance to classify the gene expression data into different classes. In this paper, followed the assumption that the gene expression data of tumor may be sampled from the data with a probability distribution on a sub-manifold of ambient space, an efficient feature extraction method named as Constrained Maximum Variance Mapping (CMVM), is presented for tumor classification. The proposed algorithm can be viewed as a linear approximation of multi-manifolds learning based approach, which takes the local geometry and manifold labels into account. The proposed CMVM method was tested on four DNA microarray datasets, and the experimental results demonstrated that it is efficient for tumor classification.

Grain distribution center is the pivot of a grain logistics system. To define the location of grain logistics distribution center is the key of grain logistics system analysis. In this paper, according to the characteristics and requirements in the selection of the location, a mathematical model applied to the location selection was established on the basis of lowest transportation cost. A hybrid ant colony algorithm was then used to solve the model, the algorithm is based on the combination of genetic algorithm and ant colony clustering algorithm. First, it adopts genetic algorithm to give information pheromone to distribute. Second, it makes use of the ant colony clustering algorithm to give the precision of the solution. The algorithm can avoid premature convergence and prevent fast local optimal solution. The instance demonstrates that the hybrid algorithm can effectively get the grain logistics center optimal solution.

This paper describes the application of our distributed computing framework for crystal structure prediction, Modified Genetic Algorithms for Crystal and Cluster Prediction (MGAC), to predict the crystal structure of the two known polymorphs of bicalutamide. The paper describes our success in finding the lower energy polymorph and the difficulties encountered in finding the second one. The results show that genetic algorithms are very effective in finding low energy crystal conformations, but unfortunately many of them are not plausible due to spurious effects introduced by the energy potential function used in the selection process. We propose to solve this by using a multi objective optimization GA approach, adding the unit cell volume as a second optimization target.

The

Orthomyxoviridae

is a family of single strained RNA viruses including five genera: Influenza virus A, Influenza virus B, Influenza virus C, Thogotovirus, and Isavirus. Usually, Influenza viruses are identified by antigenic differences in their nucleoprotein and matrix protein. In this paper, we propose an algorithm to determine a set of suitable restriction enzymes for producing recognizable restriction maps on

Orthomyxoviridae

. Our method is applied to viral strains of highly pathogenic avian influenza (HPAI), containing potentially homozygous, heterozygous, and various genetic variations. In the analysis of CAPS (Cleaved Amplified Polymorphic Sequence) subtyping, our method outperforms the RNA coding of representative and epidemiologically significant human wild-type viruses, including H3N8, H5N1, H5N9, H7N1, H7N7, and H9N2. These isolates are analyzed by CAPS with enzymes AgeI, EciI, KpnI, and XbaI. The HPAI strains show a different RFLP (Restriction Fragment Length Polymorphism) profile by comparing with other low pathogenic avian influenza (LPAI) strains. We provide a rapid, specific, and reproducible identification of the genotypes on

Orthomyxoviridae

. It permits us to quickly confirm subtypes of

Orthomyxoviridae

.

Discrimination using genetic diversity provides a significant support in genetic research and applications. Mostly, DNA markers indicate a process of determining the genotype presented at specific locations along the DNA molecule. Some developed DNA marker methods are RFLP, RAPD, AP-PCR, DAF, and AFLP. For these systems, enzymes play an important role. In this paper, we propose a mechanism to verify the enzyme efficacy for pathotyping. A procedure is given to inspect the validation on cleavage pattern by restriction enzymes, adapting the concept of genetic algorithm to quasispecies model – a genetic evolutionary processes of self-replicating macromolecules. The proposed mechanism is applied to viral strains of HPV (

Papillomaviridae

), including mutated strains from quasispecies model of homozygous, heterozygous, and various genetic variations. In the analysis of full length DNA strain PCR-RFLP subtyping, results showed that if digested patterns of HPV can be discriminated by specific enzyme set from non-high-risk and other papillomavirus, then it is also can be discriminated by the same enzyme set, under the condition of mutated simulation with quasispecies model. In addition, a measure of genetic diversity also evaluates the utility for PCR-RFLP markers in pathotyping, depending on the degree of digestion variation. We provide a specific and valid mechanism of examination on PCR-based pathotyping. Our approach offers a practical and verifiable direction for genomic pathotyping.

The Energy Distribution Markup Language (EDML) is an XML-based format that we have designed and developed to exchange protein structure energy profiles between many computers and users. Energy profiles that are distributions of various potential energies over consecutive atoms in protein molecular structures can be downloaded from the Energy Distribution Data Bank (EDB,

http://edb.aei.polsl.pl

) in the EDML format. In the paper, we describe the purpose of the EDML, a possible use of energy profiles, and internal structure of documents created in the EDML format.

In this paper, we introduce a method which uses a note model and signal post processing for a musical instrument to make a piece of music. one of the important issues in note transcription is extraction of multiple pitches. Most of the examined methods face error in joint harmonics and frequencies. A good model for note of a specified musical instrument can help us identify a note better. The presented method is based on wavelet transform, onset detection, note model and conformity reduction error algorithm or regression and post-processing for improved result. The results obtained show that detecting musical notes in a piece played on the guitar is, in comparison with similar methods, of higher detection accuracy and even in the case of noisy sound signals, the results are more acceptable.

Spectral modeling synthesis (SMS) considers a sound as a combination of a deterministic plus a stochastic component that makes possible for a synthesized sound to attain all the perceptual characteristics of the original sound. However, sometimes considerable phase variations occur in the deterministic component by using SMS since the addition of different frequency sinusoids in the overlap region causes amplitude distortion. As a result, subtraction between original and deterministic signal in time domain do not provide a good approximation of the residual signal. To overcome this problem, we propose a modified SMS that provides good approximation of the residual signal by calculating the complex residual spectrum in frequency domain. Analysis and simulation results for synthesizing bowhead whale sounds suggest that the proposed method is comparable to the SMS in both time and frequency domain. However, the proposed method outperforms the SMS in better spectrum matching because of the use of original phase information to synthesize the deterministic component as well as good approximation of the residual signal by subtracting the deterministic spectrum from the original spectrum and then utilizing spectral fitting.

Emotion helps human to express their feelings and intentions clearly. And the emphasis labels of speeches are the key of speech emotion analysis and synthesis. In order to label the emotion emphasis of speech samples from a corpus with only phonetic and prosodic information, this paper introduces an automatic labeling algorithm by measuring the prosody generation error (PGE) of the result from a statistical synthesizer. Classification and Regression Tree (CART) and Maximum Entropy (ME) modeling are adopted for automatically labeling. Experiment shows that both models are helpful for labeling.

Image feature extraction is crucial in image target recognition. This paper presents a method of image feature extraction by combining wavelet decomposition. The image is first decomposed by wavelet transforms, and the decomposed coefficients are reconstructed to form a new time series, from which some energy vector can be extracted by time-frequency domain analysis. By calculating correlation coefficients, it is possible to recognize whether target signal is involved or not in gained image. The effectiveness of the method is verified by a real image with additive simulated noise signal, especially under the condition of low SNR.

In this paper, according to characteristics of flatness error evaluation, a hybrid evaluation method to evaluate the minimum zone error is provided. The evolutional optimum model and the calculation process are introduced in detail. The hybrid optimization algorithm is based upon chaos optimization algorithm (COA) and Powell search. Compared with conventional optimum methods such as simplex search and Powell method, it can find the global optimal solution, and the precision of calculating result is very good. Moreover, the efficiency of COA is much higher than some stochastic algorithms such as simulated anneal algorithm and genetic algorithm (GA) when COA is used to a kind of continuous problems. The hybrid optimization algorithm can improve the efficiency of searching in the whole field by gradually shrinking the area of optimization variable. Finally, the control experiment results evaluated by different method such as the least square, simplex search, Powell optimum methods and GA, indicate that the proposed method does provide better accuracy on flatness error evaluation, and it has fast convergent speed as well as using computer expediently and popularizing application easily.

Crying is an acoustic event that contains information about the functioning of the central nervous system, and the analysis of the infant´s crying can be a support in the distinguishing diagnosis in cases like asphyxia and hyperbilirrubinemia. The classification of baby cry has been intended by the use of different types of neural networks and other recognition approaches. In this work we present a pattern classification algorithm based on fuzzy logic Type 2 with which the classification of infant cry is realized. Experiments as well as results are also shown.

Recent advances in physics-based sound synthesis have offered huge potential possibilities for the creation of new musical instruments. Despite that research on physics-based sound synthesis is going on for almost three decades, its higher computational complexity has limited its use in real-time applications. Conventional serial computation is inadequate for handling the physics-based sound synthesis of most instruments. To yield computation time compatible with real-time performance, we introduce a parallel approach to the physics-based sound synthesis. In this paper, with a parallel processing engine we implemented the physical modeling for one of traditional Korean plucked string instruments, called Gayageum, which has 12 silk strings. Analysis and simulation results suggest that our parallel approach has the potential to support the real-time sound synthesis of the Gayageum instrument. Moreover, our parallel approach outperforms today’s DSPs in terms of performance and energy efficiency.

Acoustic source localization has been an hot research topic with widespread applications in many fields. In the noisy environment or when the reverberation is considerable, the source localization problem becomes challenging and many existing algorithms deteriorate. The paper proposes a robust algorithm which combines the RANdom SAmple Consensus (RANSAC) algorithm, and the Generalized Cross-Correlation (GCC) based Time Difference of Arrival (TDOA). Experiments in real world data show that the proposed algorithm has significantly better performance than the traditional algorithm.

A new approach to facial expression recognition is constructed by combining the Local Binary Pattern and Laplacian Eigenmaps. Firstly, each image is transformed by an LBP operator and then divided into 3×5 non-overlapping blocks. The features of facial expression images are formed by concatenating the LBP histogram of each block. Secondly, linear graph embedding framework is used as a platform, and then Laplacian Eigenmaps is developed under this framework and applied for feature dimensionality reduction. Finally, Support Vector Machine is used to classify the seven expressions (anger, disgust, fear, happiness, neutral, sadness and surprise) on JAFFE database. The maximum facial expression recognition rate of the proposed algorithm reaches to 70.48% for person-independent recognition, which is much better than that of LBP+PCA and LBP+LDA algorithms. The experiment results prove that the facial expression recognition with local binary pattern and Laplacian Eigenmaps is an effective and feasible algorithm.

This paper primarily investigates a novel face detection method based on contourlet features. In this method, a face-pyramid is developed through contourlet transform, which includes both low and high frequency information to represent face features on multiresolutions and multidirections. The most discriminative features are then selected from the face-pyramid and are trained to construct the classifier by using the cascade boosting algorithm (Adaboost). Speed and capability are important issues for current face detection systems. This method extensively reduces feature demensions and the negative sample numbers step by step, so that the speed is increased radically. Mean-face template matching is adopted finally in the system to ensure a detection of one face in a scanned image. Extensive experiments are conducted and the results show that the proposed method is efficient in detecting frontal faces from cluttered images.

In this paper, a novel discriminant analysis named two-dimensional Heteroscedastic Discriminant Analysis (2DHDA) is presented, and used for gender classification. In 2DHDA, equal within-class covariance constraint is removed. Firstly, the criterion of 2DHDA is defined according to that of 2DLDA. Secondly, the criterion of 2DHDA, log and rearranging terms are taken, and then the optimal projection matrix is solved by gradient descent algorithm. Thirdly, face images are projected onto the optimal projection matrix, thus the 2DHDA features are extracted. Finally, Nearest Neighbor classifier is selected to perform gender classification. Experimental results show that higher recognition rate is obtained by way of 2DHDA compared with 2DLDA and HDA.

As one of the most important morphological taxonomy features, plant leaf with many strong points has significant influence on research. In this paper, we propose a novel method of plant classification from leaf image set based on wavelet transforms and support vector machines (SVMS). Firstly, the leaf images are converted into the time-frequency domain image by wavelet transforms without any further preprocessing such as image enhancement and texture thinning, and then feature extraction vector is conducted. Then the effectiveness of the proposed method is evaluated by the classification accuracy of SVM classifier. The experimental results about the data set with 300 leaf images show that the method has higher recognition rate and faster processing speed.

Dimensionality reduction is a key technology for face recognition. In this paper, we propose a novel method, called Locality Preserving Fisher Discriminant Analysis (LPFDA), which extends the original Fisher Discriminant Analysis by preserving the locality structure of the data. LPFDA can get a subspace projection matrix by solving a generalized eigenvalue problem. Several experiments are conducted to demonstrate the effectiveness and robustness of our method.

In this paper, we propose a novel approach for palmprint recognition, which combine band-limited phase-only correlation method and directional representation of palmprint. We also exploit

peak-to-sidelobe ratio

as the similarity measure. The results of experiments conducted on Hong Kong Polytechnic University Palmprint Database show that the proposed approach has higher accurate recognition rates and lower equal error rates than that of the approach, which combine band-limited phase-only correlation and original representation.

A new design foran eye gaze tracking system based on stereovision technique is presented. The system consists of two CCD cameras and two novel light sources for stereovision. The way of getting pupil position is to do subtraction of two images, the “bright pupil” and the “dark pupil”, which are gained by illuminating user’s eyes alternately. The pupil center is located by ellipse fitting when the Purkinje image is gained in the “dark pupil”, so the local gaze direction can be obtained. We also use support vector regression to figure out the mapping relationship from eye parameters to gaze point, and the interference from head motion may be eliminated by using 3D eyeball data. The experimental results show that the system can achieve an average accuracy of 1.8 degree and be robust in gaze tracking under large head movements.

Speeded-Up Robust Features (SURF), an image local feature extracting and describing method, finds and describes point correspondences between images with different viewing conditions. Despite the fact that it has recently been developed, SURF has already successfully found its applications in the area of computer vision, and was reported to be more appealing than the earlier Scale-Invariant Feature Transform (SIFT) in terms of robustness and performance. This paper presents a multi-threaded algorithm and its implementation that computes the same SURF. The algorithm parallelises several stages of computations in the original, sequential design. The main benefit brought about is the acceleration in computing the descriptor. Tests have been performed to show that the parallel SURF (P-SURF) generally shortened the computation time by a factor of 2 to 6 than the original, sequential method when running on multi-core processors.

Recently, a panoramic image has been expected in various applications due to the advantage of expressing a wide range of scenes by one image. In this paper, we propose a heuristic optimization algorithm for the panoramic image generation problem from multiple cameras. Our three-stage algorithm composed of the approximate calibration, the detailed calibration, and the image synthesis, transforms the images of the side cameras to be fit to the image of the central camera as best as possible. The image parameters are optimized by a local search method with a Tabu period as a typical heuristic optimization method. Through experiments, we show the effectiveness of our proposal.

In different methods of Biometrics, recognition by iris images in recent years has been taken into consideration by researchers as one of the common methods of identification like passwords, credit cards or keys. Iris recognition a new biometric technology has great advantages such as variability, stability and security. In this paper we propose a new feature extraction method for iris recognition based on contourlet transform. Contourlet transform captures the intrinsic geometrical structures of iris image. It decomposes the iris image into a set of directional sub-bands with texture details captured in different orientations at various scales so for reducing the feature vector dimensions we use the method for extract only significant bit and information from normalized iris images. In this method we ignore fragile bits. At last the feature vector is approximated by non linear approximation coefficient. Experimental results show that the proposed method reduces processing time and increase the classification accuracy and outperforms the wavelet based method.

Most digital cameras use a color filter array of mosaic pattern to capture the colors of the scene. To render a full-resolution color image using a singlechip camera, the missing information must be estimated from the surrounding pixels. In this paper, we present an edge-adaptive demosaicking method for color demosaicking. The algorithm first estimate missing green samples in red and blue position by determining direction of edge in horizontal, vertical and diagonal directions. After determining the green samples, missing red and blue samples estimated using color differences along the direction of edges. Experimental results demonstrate that the proposed method produces visually pleasing images and significantly outperforms existing demosaicking methods in terms peak signal-to-noise ratio (PSNR) and ΔE

ab

, which is a measure for the average color distance between original and demosaicked images in the CIELAB color space.

Direct observation using satellites and long range video surveillance is not possible for ship classification in adverse weather and during night. Radar and more specifically radar imaging offers a solution for the above adverse conditions. Ship Classification using radar is of utmost important to defense of any country to manage vast naval resources and to tell the friend from foe. Automatic ship classification based on radar images has been very successful in determining the ship class as well as other details to reliably recognize a ship type using machine vision. Inverse Synthetic Aperture Radar (ISAR) Imaging which relies on a stationary radar and a moving object with preferably superstructure will result in an image that is somewhat unique to a particular ship class. There have been many attempts to classify these ISAR images automatically with varying degree of success. The results we present here using Moment Invariants (Hu Moments) are indeed superior to many other feature-based classification approaches as they have strong invariant properties.

This paper presents an automatic machine-learning method to segment blood and bone marrow cell images. Different from traditional methods, we focus on a few significant samples rather than all of them. Firstly, three mean-shift procedures are used to seek the local clustering modes corresponding to the regions of nuclei, mature erythrocytes and background respectively. And then a SVM is trained by uniform sampling from three modes in order to find more nuclei pixels. So we could dilate the nuclei regions only in high gradient pixels to get the part pixels of cytoplasm. Finally, we train a new SVM by a training set sampling from cytoplasm and three modes to extract the whole leukocytes. SVM with fixed parameters is used here to yield two classification models via learning by sampling on-line. The segmentation results of the new method are closer to the human visual perception. It can achieve higher accuracy of segmentation in complex scenes and more robust to color confusion and changes. Experiments have demonstrated the validity of the new method compared with the thresholding and the watershed algorithm.

Recently, more attentions have been paid on finger-vein based personal identification. In real applications, finger-vein segmentation always is a crucial step to extract finger-vein features. Since finger-vein images usually are in low contrast, segmentation results often are the abridged versions of finger-vein networks. In this paper, we present a new method of finger-vein extraction based on combination of Gabor wavelets and a circular Gabor filter such that the finger-vein networks can be highlighted significantly as well as nonvascular region elimination. First, a family of Gabor wavelets is used to enhance vascular regions in an image. Then, image reconstruction is implemented using a combination rule. Finally, a circular Gabor filter is used for finger-vein extraction. Experimental results show that the proposed method is capable of extracting finger veins in an image reliably and effectively.

The quasientropy (QE) is a class of infinitely many functions of probabilities that is similar to the Shannon entropy. In this paper, we review the application of the QE approach to independent component analysis (ICA) and chaotic time series analysis. We also report the new progress of the QE approach to textural features extraction in image processing.

In this paper, we propose a new spatio-temporal representation for gait recognition. Firstly, the new representation of gait is constructed, which is the average of the Hough transformed images in one complete cycle of a silhouette sequence. Secondly, we project the new representation to low dimension by applying Principal Component Analysis. Finally, the nearest neighbor rule is adopted for recognition. The results of experiments conducted on CASIA-A Gait Database show that the proposed gait recognition approach can obtain encouraging accurate recognition rates.

This paper presents a fast algorithm for extracting features using the Simplified Gabor (SG) for Competitive Coding-based palmprint recognition. The competitive code convolves the palmprint image with a bank of Gabor filters with different orientations. We use a simplified version of Gabor filters and an efficient algorithm for extracting features to modify the competitive code. Experimental results indicate that, using SG can achieve the verification accuracy similar to using common Gabor filters, while the runtime for feature extraction using SG is very fast compare to the original algorithm.

In this paper, an effective algorithm has been proposed for palmprint recognition combining Fourier Transform and Linear Discriminant Analysis (LDA). For Fourier representation, we only exploit the center band of Fourier magnitude for recognition. The results of experiments conducted on PolyU palmprint database demonstrate the effectiveness of proposed method. All in all, the proposed method that is robust against illumination in this paper is a suitable even wonderful method for palmprint recognition.

In this paper, we present a framework to recognize books on bookshelves by reading its title on book spines. The framework consists of control and recognition module. Control module moves camera to suitable positions for image capturing while recognition one processes taken images to know which books are shelved on the shelves. Firstly, images are captured from random position. Secondly, we separate it into book and non-book regions. Then, books in book region are segmented by using line segment and MSAC based dominant vanishing point (DVP). After book verification stage, adaptive Otsu’s thresholding is employed to extract book titles and ready for recognition of next stage. In case recognizing unsuccessfully, we feedback control information to control module to adjust camera location and repeat the above procedure.

Sum of Absolute Difference (SAD) Computation is commonly used for motion estimation in video coding. It is usually the computationally intensive part in video processing. Therefore, a method to reduce the computational complexity is strictly required. In this paper, the effectiveness of saturation arithmetic on SAD computation is presented. Our goal is to use saturation arithmetic to reduce the complexity of SAD computation for the encoding process while the accuracy in finding the best matching block from the reference frame is still maintained. Experiment results show that the computational complexity of SAD computation is reduced efficiently by saving a number of bits for SAD values representation while the video quality is kept.

The traditional Hough transform needs the edge detection in advance, so the effect of edge detection influences the final fitting result. This paper proposes a new method of detecting parabolas using the kernel density estimate based on the theory of Rozenn Dahyot, and extends this method into the eyelid detection in noisy images and other images including parabolas. In our paper, the edge detection is not necessary. On one hand, we not only consider the current points on the parabola, but also ones around the parabola. Experiments demonstrate that the proposed algorithm is robust and insensitive to the noise.

IEEE recently standardized 802.21-2008 Media Independent Handover (MIH) standard. MIH is a key milestone toward the evolution of integrated heterogeneous 4G wireless networks. MIH provides number of link layer events in a unified way that facilitate upper layer protocols in making handover decisions. One such event is Link Going Down (LGD) trigger. LGD is a predictive event that is generated when link conditions are expected to degrade in near future. Traditionally such link quality degradations and connectivity losses are predicted on the basis of a single parameter only i.e. received signal strength. However, in varying wireless conditions, simple predictions relying on single link layer parameter may generate false LGD triggers. This false triggering may initiate unnecessary handovers that rather than facilitating upper layer mobility management protocols, may cause overhead and may degrade the overall network performance. In this paper, we present an intelligent model for generating MIH LGD trigger reliably. In our implementation, we used ’Time Delay Neural Networks (TDNN)’ approach using multiple link layer parameters for LGD predictions. We also analyzed the prediction accuracy and the feasibility of using such intelligent technique for mobile devices.

The ideas of using geographically distributed resources in a secure way (

Network-Internet/Grid computing

), providing self-management capabilities (

Autonomic computing

), quantifying and billing computing costs (

Utility computing

), in order to perform specific modular applications (

web services

), have been grouped altogether into the concept of

Cloud computing

.

Only commercial Cloud solutions have been implemented so far, offering computing resources and (web) services for renting. Some interesting projects, such as Nimbus, OpenNEbula, Reservoir, work on Cloud. One of their aims is to provide a Cloud infrastructure able to provide and share resources and services for scientific purposes. The encouraging results of Volunteer computing projects such as SETI@home and FOLDING@home and the great flexibility and power of the emergent Cloud technology, suggested us to address our research efforts towards a combined new computing paradigm we named

Cloud@Home

, merging the benefits and overcoming the weaknesses of both the original computing paradigms.

In this paper we present the Cloud@Home paradigm, describing its contribution to the actual state of the art on the topic of distributed and Cloud computing. We thus detail the functional architecture and the core structure implementing such a new paradigm, demonstrating how it is really possible to build up a Cloud@Home infrastructure.

Ring signcryption, a cryptographic primitive to protect security and privacy, is an encryption and authentication scheme in a single logical step which allows a user to anonymously signcrypt a plaintext on behalf of a group of users that decrypter cannot know who is the actual signcrypter, which can be used to protect nodes or participants privacy in ubiquitous environments such RFID, WSN, Ad hoc etc. In 2009, Zhang, Gao, Chen and Geng proposed a novel anonymous signcryption scheme(denoted as the ZGCG scheme) which is more efficient in computational cost and ciphertext length than the related schemes. In this paper, however, we show that the ZGCG scheme has not anonymity secure for the receiver, and then we propose an improved anonymous signcryption scheme that remedies the weakness of the ZGCG scheme. Our proposed scheme satisfies the semantic security, unforgeability, signcrypter identity’s ambiguity, and public authenticity. We also give the formal security proof in the random oracle model.

In this paper, we propose some distributed algorithms for quantized consensus. These algorithms are used to study the distributed averaging problem on arbitrary connected graphs and arbitrary connected weighted graphs, with the additional constraint that the weight value at each node is an integer. These algorithms can guarantee the system achieve consensus with some moderate assumptions and can use to solve several application problems, such as averaging in a network with finite capacity channels and load balancing in a processor network, which can be modeled as distributed averaging problem.

Grid computing provides infrastructure for solving distributed problem by sharing, selection and aggregation of distributed resources at runtime depending on their availability, performance, cost and user’s quality of service requirements. Utilization of this powerful technology is mainly conditioned by tricky management of different architectures and environments and by the difficulty to identify an efficient resource selection to map tasks into grid resources that dynamically vary their features. Resources selection needs of intelligence based automatic workflow generation to predict optimal run-time jobs allocation. In this paper we propose a dynamic job run-time scheduling system based on Java and fuzzy technology to manage Grid resources and minimize human interaction in scheduling grid jobs.

In passive RFID environment, this paper presents a pseudorandom tag arrangement for improved performance of RFID based mobile robot localization. It is assumed that a mobile robot travels along a series of linear segments, each at a constant velocity, and the number of tags sensed at one time is at most one. First, using spatial and temporal information during tag traversing, a simple but effective mobile robot localization method is developed. Second, four repetitive tag arrangements, including square, parallelogram, tilted square, and equilateral triangle, are examined. For each tag arrangement, the difficulty in tag installation and the problem of tag invisibility are discussed and compared. Third, inspired from the Sudoku puzzle, a pseudorandom tag arrangement is proposed, which is superior to conventional deterministic tag arrangement in terms of tag invisibility and tag installation.

In this paper, a novel Forward Dynamic MIMO Neural NARX model is used for simultaneously modeling and identifying both joints of the 2-axes PAM robot arm’s forward dynamic model. The contact force variation and highly nonlinear cross effect of both links of the 2-axes PAM robot arm are modeled thoroughly through a Forward Neural MIMO NARX Model-based identification process using experiment input-output training data. The results show that the novel Forward Dynamic Neural MIMO NARX Model trained by Back Propagation learning algorithm yields outstanding performance and perfect accuracy.

In this paper, a human-robot speech interface for mobile technology is described which consists of intelligent mechanisms of human identification, speech recognition, word and command recognition, command meaning and effect analysis, command safety assessment, process supervision as well as human reaction assessment. A review of selected issues is carried out with regards to recognition and evaluation of speech commands in natural language using hybrid neural networks. The paper presents experimental results of automatic recognition and evaluation of spoken commands of a manufacturing robot model simulating the execution of laser processing tasks in a virtual production process.

The paper introduces a robot visual/inertial servoing algorithm when the robot needs to track an object with inertial sensors inside. In this situation, first, inertial Jacobian is newly defined to show the relationship between an angle set velocity vector and angular velocity vector of the robot tip. That is combined with the conventional image Jacobian for the proposed robot servoing algorithm. While four landmarks have been used in the conventional visual servoing algorithm, the proposed algorithm requires only two landmarks with help of the IMU to track a moving object. Simulation and Implementation have been done to verify the feasibility of the proposed methodology.

This paper presents a service framework of a humanoid robot for the coordinated task execution. To execute given tasks, various sub-systems of the robot need to be coordinated effectively. The goal of our paper is to develop the service framework which makes it possible to execute various tasks in daily life environments. A script is used as a tool for describing tasks to easily regulate actions of the sub-systems while the robot is performing the task. The performance of the presented framework is experimentally demonstrated as follows: A humanoid robot, as the platform of the task execution, recognizes the designated object. The object pose is calculated by performing model-based object tracking using a particle filter with back projection-based sampling. An approach proposed by Kim et al. [1] is used to solve a human-like arm inverse kinematics and then the control system generates smooth trajectories for each joint of the humanoid robot. The results of our implementation show the robot can execute the task efficiently in human workspaces, such as an office or home.

We propose an efficient and powerful alternative for adaptation of human motions to humanoid robots keeping the bipedal stability. For achieving a stable and robust whole body motion of humanoid robots, we design a biologically inspired control framework based on neural oscillators. Entrainments of neural oscillators play a key role to adapt the nervous system to the natural frequency of the interacted environments, which show superior features when coupled with virtual components. The coupled system allows an unstable system to stably move according to environmental changes. Hence the feature of the coupled system can be exploited for sustaining the bipedal stability of humanoid robots. Also based on this, a marionette-type motion conversion method to adapt captured motions to a humanoid robot is developed owing that there are the differences in capabilities of dynamics and kinematics between a robot and a human. Then this paper discuss on how to stably show human motions with a humanoid robot. We verify that a real humanoid robot can successfully sustain the bipedal stability exhibiting captured whole body motions from various simulations and experiments.

The systems let user track their eye gaze information have been technologically possible for several decades. However, they are still very expensive. They have limited use of eye tracking and blink detection infra-structure. The purpose of this paper is to evaluate cost effects in the sector and explain our new approach in detail which reduces high costs of current systems apparently. This paper introduces an algorithm for fast and sub-pixel precise detection of eye blobs for extracting eye features. The algorithm is based on differential geometry and still exists in OpenCpV library as a class. Hence, blobs of arbitrary size that means eye size can be extracted by just adjusting the scale parameter in the class function. In addition, center point and boundary of an eye blob, also are extracted. These describe the specific eye location in the face boundary to run several algorithms to find the eye-ball location with its central coordinates. Several examples on real simple web-cam images illustrate the performance of the proposed algorithm and yield an efficient result on the idea of low-cost eye tracking, blink detection and drowsiness detection system.

In this paper, we present a two-stage vision-based approach to detect front and rear vehicle views in road scene images using eigenspace and a support vector machine for classification. The first stage is hypothesis generation (HG), in which potential vehicles are hypothesized. During the hypothesis generation step, we use a vertical, horizontal edge map to create potential regions where vehicles may be present. In the second stage verification (HV) step, all hypotheses are verified by using a Principle Component Analysis (PCA) for feature extraction and a Support Vector Machine (SVM) for classification, which is robust for both front and rear vehicle view detection problems. Our methods have been tested on different real road images and show very good performance.

In different areas of Biometrics, recognition by iris images in nowadays has been taken into consideration by researchers as one of the common methods of identification like passwords, credit cards or keys. Iris recognition a novel biometric technology has great advantages such as variability, stability and security. Although the area of the iris is small it has enormous pattern variability which makes it unique for every one and hence leads to high reliability. In this paper we propose a new feature extraction method for iris recognition based on contourlet transform. Contourlet transform captures the intrinsic geometrical structures of iris image. It decomposes the iris image into a set of directional sub-bands with texture details captured in different orientations at various scales so for reducing the feature vector dimensions we use the method for extract only significant bit and information from normalized iris images. In this method we ignore fragile bits. At last, the feature vector is created by using Co-occurrence matrix properties. For analyzing the desired performance of our proposed method, we use the CASIA dataset, which is comprised of 108 classes with 7 images in each class and each class represented a person. And finally we use SVM and KNN classifier for approximating the amount of people identification in our proposed system. Experimental results show that the proposed increase the classification accuracy and also the iris feature vector length is much smaller versus the other methods.

In this paper, an algorithm for vehicle license plate detection (VLPD) is proposed, to select automatically statistical threshold value in HSI color space. The proposed VLPD algorithm consists of two main stages. Initially, HSI color space is adopted for detecting candidate regions. According to different colored LP, these candidate regions may include LP regions; geometrical properties of LP are then used for classification. The proposed method is able to deal with candidate regions under independent orientation and scale of the plate. Finally, the decomposition of candidate regions contains predetermined LP alphanumeric characters by using position in the histogram to verify and detect vehicle license plate (VLP) region. In experiment more than 150 images were used, they were taken from the variety of conditions such as complex scenes, illumination changing, distances and varied weather etc. Under these conditions, success of LP detection has reached to more than 94%.

In buildings, a practical sensing system to collect occupant location information has great importance in improving occupants’ comfort and utilizing energy more efficiently by optimizing control strategies of lighting, HVAC devices and elevators. We implement a practical algorithm for occupant detection in use of existing video camera hardware. In our system, we present a novel blob segmentation method based on rule and propose a fast template-based head detection algorithm that matches directly on gradient maps other than edge maps. The accuracy is improved and can satisfy the need of the control system in smart buildings. The speed is about twice as fast as traditional algorithms.

Carrying out user commands entails target object detection for service robots. When the robot system suffers from a limited object detection capability, effective communication between the user and the robot facilitates the reference resolution. We aim to develop a service robot, assisting handicapped and elderly people, where most of the user requests are directly or indirectly linked to some objects in the scene. Objects can be described using features like color, shape, size etc. For simple objects on simple backgrounds, theses attributes can be determined with satisfactory results. For complex scenes, position of an object and spatial relation with other objects in the scene, facilitate target object detection. This paper proposes a spatial relation model for the robot to interpret user’s spatial relation descriptions. The robot can detect a target object by asking the user the spatial relationship of the object and some known objects automatically recognized.

This paper describes an approach to extract windows by analyzing geometrical characteristics of building surface. Firstly, building surfaces are detected and then wall region is extracted by using hue color of pixel; this step was well described in our previous works. The non-wall regions are considered as candidates of other components of building such as windows, doors, columns and so on. To extract the windows, the image of candidates is recovered in rectangular shape. Then the ambiguous candidates which have irregular shape, for example, long and thin or very small are coarsely rejected. The geometrical characteristics such as the center coordinates, area, aspect ratio and the aligned coexistence are used for extracting the windows. The proposed approach has been experimented for a database with 150 building surfaces comprising 1607 windows. We obtained 93.34% extraction rate.

This paper describes an auto-surveillance system which tracks a person who comes in/out an office using multiple camera system. Furthermore it automatically recognize whether the person bring an object in/out. For this purpose, we set three steps. The first step is detecting a person using MBM(Multiple Background Model) and TMB(Temporal Median Background). The second step is calculation of correspondence between persons detected by different view-point cameras in the multiple camera system. We simply calculate the correspondence based on the principal axis and homography. The last step is generating global color model, which includes every local color model organized by GMM (Gaussian Mixture Model) from each camera, of the person. The global color model represented by GMM checks the temporally varied error and detects the object to bring in or out objects. In the experiment, we show the detected human silhouette by background subtraction and the tracking result by correspondence of multiple views. We also show the color segmentation using GMM and the recognition result for detecting objects brought in/out by the tracked person.

This paper describes the method to know objects for autonomous robot navigation in an unknown outdoor environment. The method segments the objects from an image taken by moving robot on outdoor environment. In the beginning object segmentation, this uses multiple features to obtain the objects of segmented region. Multiple features are color, context information, line segments, edge, Hue Co-occurrence Matrix (HCM), Principal Components (PCs) and Vanishing Points (VPs). The model of the objects for outdoor environment defines their characteristics individually. We segment the region as mixture using the proposed features and methods. Next the stage classifies the object into natural and artificial ones. We detect sky and trees of natural object and detect building of artificial object using the combination of appearance and context information. Then we estimate the dimensions of building. Extensive experiments with the object segmentation and analysis on outdoor environment confirm the validity of the approach.

In order to get the features of moving vessels at the port correctly and track the target quickly and efficiently, we combined the advantages of traditional invariant moments and invariant line moments, and proposed a object recognition algorithm based on pseudo invariant line moments. Using this algorithm, first we get the calculation regions of the objects in an image, then do edge detection to the calculation regions and get the pseudo invariant line moments by calculating binary image. The experimental results show that the algorithm can not only get the regions of moving objects quickly and accurately, but also can predict the directions of moving objects effectively. This algorithm is applied in the intelligent video monitoring system of moving vessels successfully.

The most important things to realize such an intelligent system are core functions such as landmark detection, recognition and reconstruction. Since where we have core functions, the intelligent system can propagate other procedures like navigation, mapping, localization, etc. Thus, this paper describes an approach to construct a structural data for core functions by using geometrical structure of building. Firstly, line segments are detected. Then several processes such as rejecting noises, calculating dominant vanishing points, filtering the edges of building are used to detect the building surfaces. The criteria are created for decision of building detection function. Secondly, for each surface, a generative model including area, wall histogram and a list of local features are computed for the recognition function. Finally, the geometrical features as windows, doors, floors or rooms are estimated for reconstructing the building. The proposed method has been performed with large databases and sound results of all functions.

In this paper, a method is proposed, to solve correspondence problem under structured space which is installed multiple cameras. The correspondence between different cameras is an important task to use the multiple camera system. For solving this problem, the proposed method is consists of three steps which are detection of moving object, feature extraction and correspondence among different cameras. First step is to detect moving people by background subtraction from multiple background model. The temporal difference is used jointly to remove noise occurred from temporary change. The detected regions are divided using labeling as individual person. The second step is to segment the each person by a criterion with appearance and context information. The segmented regions in a person are estimated as Gaussian mixture model (GMM) for correspondence. The final step is process of correspondence between different cameras. A GMM from a camera is matched with another GMM from other cameras. A ratio of those GMMs is used as a criteria to identify same person. The experiment was performed with the specific scenarios in quantitative results.

For integrated social networking and sensor networking services, a unified approach using a unified directory service based on web-based directory was studied. As a convenient and usable mobile web service for unified social/sensor networking service, the multi-lingual single-character domain names as mobile user interface for accessing the metadata of social/sensor information in unified directory are convenient, efficient and consistent. For searching for social/sensor information as well as registering metadata of sensor/social information, we introduce the web-based unified-directory service with the requirements, performance metrics for QoS, resource utilization and real-time estimation of the performance metrics.

Energy markets are systems for effecting the purchase and sale of electricity using supply and demand to set the price. A typical energy market involves a wholesale market for electricity generation, when competing generators offer their electricity output to retailers, and a retail market for electricity retailing, when end-use customers choose their supplier from competing electricity retailers. This paper addresses the challenges created by competitive energy markets towards ensuring the full benefits of deregulation. It presents a multi-agent energy market composed of multiple autonomous computational agents, each responsible for one or more market functions, and each interacting with other agents in the execution of their responsibilities. Additionally, the paper presents a negotiation model for autonomous agents. The model handles bilateral multi-issue negotiation and formalizes a set of negotiation strategies studied in the social sciences and frequently used by human negotiators.

The paper presents an approach to automated user interest matching in online classified advertising systems which is based on the analysis of the structure and semantics of classified ads. A classified advertisement structure and classified ad types along with the examples are described in the paper.

There are several existing buyer coalition schemes. These schemes do not consider forming a buyer coalition with bundles of items. There is only one scheme that forms a buyer coalition with bundles of items. Nevertheless, the scheme suffers from computational complexity. In this paper, we have applied genetic algorithms (GA) to form buyer coalitions with bundles of items, called the GroupPackageString scheme. The fitness function is defined by total discount of the buyer coalitions over the GA to measures the GroupPackageString scheme. The coalition results show that the total discount of any coalition in this scheme is higher than those in the GroupBuyPackage scheme.

In wireless sensor networks (WSNs) individual sensor nodes are subject to security compromises. An adversary can use compromised sensor nodes to inject false reports into the WSN. If undetected, these false reports are forwarded to the base station. Such attacks by compromised sensor nodes can not only result in false alarms but also depletion of the limited amount of energy in battery powered sensor nodes. The statistical en-routing filtering (SEF) scheme can detect and drop false reports during the forwarding process. In SEF, the number of the message authentication codes (MAC length) is important for detecting false reports and saving energy in network. In this paper, we present a fuzzy-based MAC length determination method for SEF. If there are fewer nodes surrounding the occurred event in the field in the network than the MAC length, the node cannot generate a legitimate report in SEF. The fuzzy-based method can prevent this problem and provide energy savings. We evaluated the proposed method’s performance via simulation.

The issue of identifying appropriate cluster-heads has recently been the focus of extensive research and development in wireless sensor networks. Unfortunately, cluster-heads are generally chosen either in a random manner or mainly based on nodes’ residual energy. Accordingly, there is no guarantee that network coverage is well-preserved while this QoS is vital in target tracking and surveillance applications. In order to enhance both coverage preservation and energy efficiency, we propose a Coverage and Energy Aware Cluster-Head Selection Algorithm which fully considers three critical factors: the node’s energy, location and especially coverage cost metric. Simulation results demonstrate that our algorithm cannot only prolong the network lifetime over 11%, but also substantially enlarge network coverage, from the middle phase of the network lifetime, by over 20% compared to the traditional energy-based selection methods in LEACH and HYENAS system.

We developed a middleware platform, i.e. COSMOS (Common System for Middleware of Sensor Network), for several types of sensor networks including the Zigbee wireless sensor network, the CDMA cellular phone-based network, the RFID reader-based network and the IP-USN based on 6LowPAN. Development has been focused on interfaces for application programs as well as on sensor network abstractions for various ubiquitous sensor networks (USN). Standard interfaces were defined between the USN middleware and USN networks as well as application services. We studied several USN services including u-Healthcare to examine important issues about middleware platform for integrating with other standardized communication framework, e.g. a medical device communication framework. We introduce application services and the real-time data analysis for QoS in the u-healthcare service.

In order to effectively monitor the state of software running, it is necessary to embed the software sensor into the program. New software sensors based on aspect oriented technology was designed for overcoming the deficiencies of traditional method in software sensor design and implantation. An inertial navigation system was taken as example and discussed. Practical application shows that the method of software design and implantation of the sensor has guiding significance.

Based on the vibration spectrum analysis, this paper proposed a CMAC (Cerebellar Model Articulation Controller) neural network diagnosis technique to diagnose the fault type of turbine-generator sets. This novel fault diagnosis methodology contains an input layer, quantization layer, binary coding layer, excited memory addresses coding unit, and an output layer to indicate the fault type possibility. Firstly, we constructed the configuration of diagnosis scheme depending on the vibration fault patterns. Secondly, the known fault patterns were used to train the neural network. Finally, combined with a Visual C++ program the trained neural network can be used to diagnose the possible fault types of turbine-generator sets. Moreover, a PIC microcontroller based portable diagnosis apparatus is developed to implement the diagnosis scheme. All the diagnosis results demonstrate the following merits are obtained at least: 1) High learning and diagnosis speed. 2) High noise rejection ability. 3) Eliminate the weights interference between different fault type patterns. 4) Memory size is reduced by new excited addresses coding technique. 5) Implement easily by chip design technology.

This paper proposed a new fault diagnosis method based on the extension theory for analog circuits. The responses of an analog circuit were difference at some node with the normal and failure conditions. However, the identification of the faulted location was not easily task due to the variability of circuit components. So this paper presented a novel extension method for fault diagnosis of analog circuit, which is based on the matter-element model and extended relation functions. The proposed method has been tested on a practical analog circuit, and compared with the multilayer neural network (MNN) based methods and k-means classification method. The application of this new method to some testing cases has given promising results.

To improve the structure of the twin seat underframe in multiple unit train and lightweight design, FEM is used to analyze this problem. First of all, the 3D geometric model of the existing twin seat underframe is built. And according to the requirements of enterprises and experiment, the finite element model is built and analyzed. Then results of FEA and experiment are compared. According to the comparison, the finite element model that was simplified is effective. On the basis of effective finite element model, five schemes of the twin seat underframe are proposed and structural strength of twin seat underframes that were improved is researched. Finally according to influence on manufacturing procedure of the twin seat underframe and the error of the FEA, the ideal solution is proposed.

In this article, an LED Array Color Temperature Controlled Lighting System has been implemented using an 8 bit RISC IP Core for the lighting control system, as well as a Color Temperature Controlling IP and Delta-Sigma DAC IP designed to control the system. The light sources are made of an LED Array, and the LEDs are configured to have 10 stick bars, such as 3 chips of white (30EA), daylight (30EA), red (30EA), green (30EA), and blue (30EA) 0.1W SMD. The time information is acquired through a Real Time Clock, and bio rhythm compatible presentation is made through the LED Array Color Temperature control. The temperature control IP and Delta-Sigma DAC IP are interfaced by accessing the SFR address of the 8 bit RISC IP Core. The system is configured so that the Delta-Sigma DAC IP would produce 0V~3.3V analogue signals through a low bandwidth passing filter and control the lighting system through the serial communications with a PC using the serial port.

This paper introduces a universal data access (DA) server for modern distributed data acquisition and monitoring systems used in process and factory automation. This system is proposed with utilization of the OPC (Openness, Productivity, and Collaboration) technology and XML to achieving interoperability and platform independence. It allows to easily aggregate a large number of existing OPC DA servers and new OPC XML-DA servers into a unified and flexible system that supports exchange of data among these servers. By using binary data encoding to the SOAP messages, the proposed system has a sufficient good performance. The security consideration is discussed to provide more information to technical-level readers. The comparison of the proposed system with the existing approaches is also presented.

In this paper, a Petri net-based logic and sequence control model of photo mask transport was constructed. A characteristic analysis was made on the established Petri net model in behavior and structure. This checked the safeness and reliability of the constructed control model off line. In order to implement the Petri net-based control model of photo mask transport via a Programmable Logic controller (PLC), an approach was proposed to generate Ladder Logic Diagram (LLD) programs from the Petri net-based control model. Execution of the generated LLD programs validated the effectiveness of the proposed method.

The purpose of this study is to design a tracking controller for micro-piezoelectric motion platform applications. The hysteresis effect is originated from the piezoelectric actuated platform that provides nonlinear behaviors. A Prandtl-Ishlinskii model is constructed to describe the hysteresis behavior of piezoelectric actuators. The weights of hysteresis model are identified by using the LMS(Least-Mean-Square) algorithm. Based on the Prandtl-Ishlinskii model, a feed-forward controller is developed for compensating the hysteresis nonlinearity. A self-tuning neuro-PID controller is introduced to suppress the tracking errors due to the modeling inaccuracy and hence to get precision tracking errors. These approaches are numerically and experimentally verified which demonstrate the performance and applicability of the proposed designs under a variety of operating conditions.

The DSTATCOM regulates the voltage at the point of common coupling (PCC) by injecting reactive power to the PCC, making it meet the requirement of the voltage quality. First, this paper deduces from the balance theory of instantaneous power of the DSTATCOM system the direct output voltage control strategy, in which the current detection circuit is not demanded. Compared with the cascade control strategy, it has the merits of a simple structure and fast response, but it doesn’t perform well when the parameters of the system are changed. So the paper proposes the fuzzy PI control to solve the problem. The validity and effectiveness of the control strategy has been verified by theoretical analysis and digital simulation.

The Brushless DC Motor (BLDC) has been applied widely for its high torque density, high efficiency and small size, but its torque ripple is relatively high. In recent years, some scholars apply direct torque control to BLDC, for reducing torque ripple by means of the fast response of torque. Based on this theory, this paper presents a complex fuzzy controller which comprises two fuzzy controllers. The first controller is designed to select correct voltage vector according to the torque error, stator flux-linkage error and electric angle of stator flux-linkage. The second fuzzy controller with adjustable factor is designed to regulate the action time of voltage vector according to the torque error and torque error differential. For minimizing the torque ripple, the genetic algorithm (GA) is utilized to optimize adjustable factor. The whole system is simple, the control method is convenient to be realized and the effect is significant. Simulation and experiment results verify the effectiveness of the complex fuzzy controller.

In this paper we address scheduling of the permutation flow shop with minimization of makespan and total flow time as the objectives. We propose a memetic algorithm (MA) to search for the set of non-dominated solutions (the Pareto optimal solutions). The proposed MA adopts the permutation-based encoding and the fitness assignment mechanism of NSGA-II. The main feature is the introduction of an NEH-based neighborhood function into the local search procedure. We also adjust the size of the neighborhood dynamically during the execution of the MA to strike a balance between exploration and exploitation. Forty public benchmark problem instances are used to compare the performance of our MA with that of twenty-seven existing algorithms. Our MA provides close performance for small-scale instances and much better performance for large-scale instances. It also updates more than 90% of the net set of non-dominated solutions for the large-scale instances.

In this article, a robust nonlinear friction control strategy is developed using friction observer and recurrent fuzzy neural network. The adaptive dynamic friction observer based on the LuGre friction model is proposed to estimates the friction parameters and a directly immeasurable friction state variable. A RFNN approximator and reconstructed error compensator is also designed to give additional robustness to the control system due to the presence of the friction model uncertainty. A proposed composite control scheme with basic basckstepping controller is applied to the position tracking control of the servo mechanical system.

Teams of unmanned combat aerial vehicles (UCAVs) are well suited to perform cooperative mission in hostile environment, and cooperative path planning holds great attention for improving the efficiency of multi-UCAV combating. In this paper, a mathematical formulation for cooperative path planning problem is presented based on the analysis of typical constraints in the scenario. Different from previous studies, the formulation introduces cooperation coefficient to estimate how the UCAV flight paths fulfill the cooperative constraints. Then a coevolutionary multi-ant-colony algorithm is designed and implemented to solve the above-mentioned problem, based on multi-ant-colony algorithm and coevolutionary strategy. The state transition rule and pheromone updating strategy is modified to increase the algorithm performance. Finally, the proposed method is validated to be effective and feasible to solve the cooperative constraints efficiently, and is effective for the multi-UCAV cooperative path planning problem.

This paper deals with an autonomous flight algorithm design problem for the tilt-rotor airplane under development by Korea Aerospace Research Institute for simulation study. The objective of this paper is to design a guidance and control algorithm to follow the given command precisely. The approach to this objective is that model-based inversion is applied to the highly nonlinear tilt-rotor dynamics at fixed-wing mode (nacelle angle=0 deg), and then the classical controller is designed to satisfy overall system stabilization and precise command following performance. Especially, model uncertainties due to the tilt-rotor model itself and inversion process are adaptively compensated for in a simple neural network for performance robustness. The designed algorithm is evaluated from the nonlinear airplane simulation in fixed-wing mode to analyze the command following performance for given trajectory. The simulation results show that the command following performance is satisfactory and control responses are within control limits without saturation.

This paper works toward the development and implementation of a INS/GPS integration system for the land vehicle application. A developed INS system is introduced to keep measuring the position/orientation of the vehicle when the vehicle is passed through GPS signal shading area. A new orientation scheme is studied to full fill the measurement states of the integration system. Roll/pitch estimation compensating external acceleration is performed with inertial sensors and yaw angle is obtained with GPS information. And then, the orientation information is supplied to the linearized Kalman filter of error model. This process is shown to improve the performance of the integration system. The field test was performed along a non-flat contour with some dismissals of GPS on it.

Of particular interest is intrinsic principle for synchronization behavior of complex networks. The synchronization behaviors of coupled Lorenz chaos systems via three kinds of networks are explored, respectively. The bounded property and synchronization criteria for complex networks under consideration are derived. The intrinsic principle for synchronization behavior depends on synchronization error matrix and coupling strength.

A fast iteration algorithm proposed is applied to solve the latitude of geocentric earth coordination from geocentric Cartesian coordination. Comparing with other well-known algorithms, the simulation results demonstrate that the algorithm is better in computation speed, computation precision and simpler in computation complexity; therefore, it is available for the portable.

On one hand,it has been demonstrated theoretically and verified numerically that neural networks can act as a time domain nonparametric modeling approach of engineering dynamic systems by forecasting their dynamic responses according to them in the past consequent time steps. On the other hand, as a time-domain auto-regressive method, the auto-regressive and moving average (ARMA) model has been widely employed to describe the mapping between structural dynamics response at a current time instant and them in the past previous time instants. The equivalency of the physical meaning of the neural network nonparametric model and the ARMA parametric model for dynamic systems is testified firstly in this paper. Then, a novel structural parametric identification methodology based on the nonparametric neural network model is proposed by the use of excitation and dynamic response measurement time series. The accuracy and efficacy of the proposed strategy for a multi-storey frame structure model are validated using the excitation and acceleration measurement time series of impact test.

In this paper, we present an application of neural networks in the renewable energy domain. We have developed a methodology for the daily prediction of global solar radiation on a horizontal surface. We use an ad-hoc time series preprocessing and a Multi-Layer Perceptron (MLP) in order to predict solar radiation at daily horizon. First results are promising with nRMSE < 21% and RMSE < 998 Wh/m

2

. Our optimized MLP presents prediction similar to or even better than conventional methods. Moreover we found that our data preprocessing approach can reduce significantly forecasting errors.

This paper proposes the use of non-extensive entropy for text classification. Non-extensive entropy technique is used for text classification by estimating the conditional distribution of the class variable given the document. The underlying principle of non-extensive entropy is that without external knowledge, one should prefer distributions that are uniform. This paper proposes two models for text classification based on maximum entropy principle. The first model extends Shannon entropy into non-extensive entropy to simplify the form of classifier, the other one introduces high-level constraints into non-extensive model to impose constraints on the pairs of entities. Model with high-level constraints constructs relations between word pairs which builds semantic constraints, for the sake of advancing accuracy of text classification. Experiments on the 20_newsgroup set demonstrate the advantage of non-extensive model and non-extensive model with high-level constraints.

This paper presents a new concept of intelligent e-learning systems with intelligent two-way speech communication between an e-learning system and the user. The system uses intelligent methods for analysis, evaluation and assessment of user knowledge and skills as well as e-learning process control, supervision and optimization. Developed as a prototype for mobile technologies, the communication system by speech and a natural language between the intelligent e-learning system and external users consists of intelligent mechanisms for biometric user identification, speech recognition, word and sentence recognition, sentence meaning analysis, and user reaction assessment. Also discussed are selected problems of the new concept of intelligent e-learning systems using intelligent speech communication. The discussion focuses on recognition and evaluation of spoken natural language sentences with use of hybrid neural networks.

KMKE provides a knowledge engineering approach to integrating knowledge management activities (such as knowledge modeling, knowledge verification, knowledge storage and knowledge querying) into a systematic framework. In this paper, we develop the KMKE knowledge management system based on design patterns and parallel processing. First, several design patterns are applied to develop the KMKE system for enhancing its flexibility and extensibility. Making the KMKE system flexible and extensible is useful to deal with continuous changes originated in knowledge. Second, JAVA programs and CLIPS programs are bound to offer the capability of knowledge inference for the KMKE system. Knowledge verification and knowledge querying can then be performed through the execution of CLIPS rules. Finally, we propose the Parallel CLIPS to shorten the execution time of the KMKE system. Since a large amount of knowledge may increase the execution time substantially, parallelizing the execution of CLIPS rules in cluster system could effectively reduce the search space of the CLIPS inference engine.

Nowadays, due to the spread of digital imaging technologies, the design of effective content based image retrieval (CBIR) systems is perceived by the research community as a primary problem. Various techniques such as clustering and relevance feedback were proposed to obtain a certain level of knowledge about a given image database. Often clustering techniques were used to obtain a first level characterization of the image database used to speed up the successive stage of queries. In this work the authors use the knowledge obtained using a fuzzy clustering algorithm to reinforce the user feedback. The system was tested on the Columbia Coil-20 image database and the obtained results seem to be encouraging.

Plant has plenty use in foodstuff, medicine and industry, and is also vitally important for environmental protection. So, it is important and urgent to recognize and classify plant species. Plant classification based on leaf images is a basic research of botanical area and agricultural production. Due to the high nature complexity and high dimensionality of leaf image data, dimensional reduction algorithms are useful and necessary for such type of data analysis, since it can facilitate fast classifying plants, and understanding and managing plant leaf features. Supervised locally linear embedding (SLLE) is a powerful feature extraction method, which can yield very promising recognition results when coupled with some simple classifiers. In this paper, a semi-SLLE is proposed and is applied to plant classification based on leaf images. The experiment results show that the proposed algorithm performs very well on leaf image data which exhibits a manifold structure.

Disease Candidate Genes (DCGene) is an advanced system for predicting the disease related genes, It is a novel computational approach by using the GO annotation information. The performance of the DCGene is evaluated in a set containing 1057 test samples, on both the local region and genome scale. In the local region scale, for 397 of 1057 (37.6%) samples, the disease-associated genes are at the top 1 of the out put gene prioritization list, and if the top 9 genes are all considered, 754(71.3%) disease-associated genes are included in the result. In the genome scale, 55% of the disease-relevant genes are included in the top scoring 3%, and 74% of the disease-relevant genes are included in the top 15%. The performance of the DCGene is demonstrated to be significant better than the others by comparison with the other systems and methods.

This paper describes a complete image processing framework for Virtual Colonscopy. The developed algorithms cover the entire process that allows a virtual navigation inside the colon lumen, starting from a dataset of axial CT slices. The implemented modules are: electronic colon cleansing, lumen segmentation, skeletonization, rendering and navigation. In particular for the centerline problem two different techniques are proposed and evaluated.

This paper describes a CAD system to detect a particular type of colon cancer, flat polyps. The identification of suspicious regions is based on two type of analysis executed in succession: a densitometry analysis that researches contrast fluid on polyp surface and a morphological analysis that reduces number of false positives, calculating a curvature index of the surface.

Virtual Colonoscopy is an innovative method to discover colon neoplasias created in order to alleviate patients aches generated by the standard colonoscopy. For the same reason, we have realized an automatic process finalized to find polyps into the lumen through the extraction of colon centerline and the calculation of polyps distance from anus. This paper contains the description of what is implemented. In particular, the developed algorithms build up following steps: colon lumen segmentation starting from a dataset of axial CT slices, 3D rendering, centerline extraction and evaluation of polyps distance from anus.

This paper presents a new approach for the classification of pathological vs. healthy endomicroscopic images of the alveoli. These images, never seen before, require an adequate description. We investigate two types of feature vector for discrimination: a high-level feature vector based on visual analysis of the images, and a pixel-based, generic feature vector, based on Local Binary Patterns (LBP). Both feature sets are evaluated on state-of-the-art classifiers and an intensive study of the LBP parameters is conducted. Indeed best results are obtained with the LBP-based approach, with correct classification rates reaching up to 91.73% and 97.85% for non-smoking and smoking groups, respectively. Even though tests on extended databases are needed, first results are very encouraging for this difficult task of classifying endomicroscopic images of the distal lung.

Distributed denial of service (DDoS) attack is one of the major threats to the current Internet. The IP Flow feature value (FFV) algorithm is proposed based on the essential features of DDoS attacks, such as the abrupt traffic change, flow dissymmetry, distributed source IP addresses and concentrated target IP addresses. Using linear prediction technique, a simple and efficient ARMA prediction model is established for normal network flow. Then a DDoS attack detection scheme based on anomaly detection techniques and linear prediction model (DDAP) is designed. Furthermore, an alert evaluation mechanism is developed to reduce the false positives due to prediction error and flow noise. The experiment results demonstrate that DDAP is an efficient DDoS attacks detection scheme with more accuracy and lower false alarm rate.

In this paper, we propose a new modified version of Advanced Encryption Standard (AES) using chaotic key generator for satellite imagery security. We analyze and examine the Modified AES and chaotic key generator to enhance the key space and sensitivity, performance, and security level for reducing the risk of different attacks. The chaotic key generator utilizes multiple chaotic maps named as Logistic, Henon, Tent, Cubic, Sine and Chebyshev. The proposed algorithm presents numerous interesting and attractive features, such as a high level of security, large enough key-space with improved key sensitivity, pixel distributing uniformity and an acceptable encryption and decryption speed. The presented algorithm is ideal for real-time applications to deal with redundant, bulky, complex and stubborn satellite imagery.

Recently, proxy re-encryption scheme received much attention. A proxy re-encryption used for divert ciphertext from one group to another without revealing underlying plaintext is proposed in this paper. The scheme is bidirectional and any member can independently decrypt the ciphertext encrypted to its group. The security of the proposed scheme is discussed and proofs are given to show that the scheme withstands chosen ciphertext attack in standard model.

It is one of the serious challenges in bioinformatics that the algorithms are used to study cell’s molecular pathways and to simulate cell’s molecular network so as to reveal tumor’s molecular characteristics at a molecular level. In this paper we aim at disclosing gene nonlinear interactions under parallel computing environment. First based on graph-coloring scheme, determine the types of the higher order logical relationship among multi-genes to get their expression pattern. Secondly gets the sample supporting degree for the logical expression patterns. Thirdly take the supporting degree for the weight of the regulatory network model to show the probability with which the logical relation happens among these samples, and further build a weighted and directed regulatory network of gene expression. Finally apply this method to the colon cancer mRNA micro-array dataset to build a higher order logical regulatory network and to visualize the tumors’ molecular signal pathways. Results show that with this way we can not only extract multi-gene’s nonlinear logical relations hidden in the gene expression profile but also analyze effectively tumor cell’s signal pathways and build a regulatory network of gene expression, which provides a tool for the study on tumor-gene’s molecular bioinformatics.

The broad knowledge source in the agricultural field causes many problems such as poor knowledge structure, fuzzy and uncertain representation of objective phenomena, which requires that, in the agricultural intelligent system, the knowledge representation and processing pattern could reflect this kind of uncertainty or fuzziness. The representation and reasoning capability of traditional production rules, however, is somewhat insufficient in the representation of knowledge uncertainty or fuzziness. In order to overcome the foregoing insufficiency, the weighed fuzzy logic production rule was put forward to characterize the uncertainty or fuzzy knowledge; the descriptive method of fuzzy production rules was proposed based on BNF, finally, the feasibility and validity of fuzzy production rules on the representation of the uncertain and fuzzy agricultural knowledge was tested with the implemented instance of wheat expert system.

Radiotherapy can induce DNA damage into cells and trigger the cell cycle arrest and cell apoptosis by regulating the vital genes and their signal pathways. To illustrate the cellular self-defense mechanisms in fighting against genome stresses under radiotherapy, a model of P53 stress response networks is proposed by using the methods with the system biology and cyber-biology at single cell level. The kinetics of Double Strand Breaks (DSBs) generation and repair, ARF and ATM activation, P53-MDM2 feedback regulation, as well as the toxins degradation are presented versus continuous radiation time. The model provides a theoretical framework to illustrate the complicated kinetics in cellular response to acute IR under radiotherapy.

In response to genome stresses, cell can trigger its self-defensive mechanism by regulating the vital genes and their complicated signal pathways. To illustrate the cellular response DNA damage under radiotherapy, a plausible feedback-control model of P53 stress response networks is proposed at single cell level. The kinetics of double strand breaks (DSBs) generation and repair, ARF and ATM activation, P53-MDM2 regulation, toxins degradation, as well as ion radiation (IR) dose feedback-control are presented.

Objective: To investigate the dependence of insulin sensitivity index (SI) and glucose effectiveness (SG) on the parameters of insulin kinetics in Bergman’s minimal model. Methods: The concentration values of glucose and insulin at series of time were computated, where the parameters in the model were generated randomly in a certain range. Based on the optimization method, insulin sensitivity index and glucose effectiveness were estimated. Results: The largest relative error of SI was less than 1.66%

0

and the largest relative error of SG was less than 0.96%

0

, if a data set of 240 points of concentration value was used in computation. However, if there are only 27 points (according to intra venous glucose tolerance test) in the data set, the relative error of SI ranged from 0.45% to 48.1%, and the relative error of SG was less than 5.6%. Conclusion: The dependence of SI and SG upon parameters of insulin kinetics is not significant.

In the process of embedded systems modeling and performance evaluation, the performance indicators, such as average delay and CPU occupancy rate, are often need to be analyzed and evaluated. In this paper, the method for performance evaluation has been proposed, and the library function for performance evaluation based on SystemC language has been designed. The library function is designed for evaluating performance parameters, calculating expectations and confidence interval, and terminating the simulation when the performance indicators meet the accuracy requirements. In this paper, the performance evaluation of embedded speech recognition systems by using the library function and the method proposed was introduced in detail. For validating the correctness and practicality of the library function further, the library function has been also used in the performance evaluation of wireless AP systems, as well as in performance evaluation of complex embedded systems based on SystemC. It is of great value for extending the library of SystemC.

The objective of this paper is to propose an intelligent computing algorithm, represented by an artificial neural network model, to analyze the dynamics of stock prices of banks. Through the empirical application of the model developed, it is expected to obtain indications about the ability of the artificial neural network model developed to generalize the phenomenon analyzed. So the research aims to provide empirical results about the use of non-linear methods of analysis for the study of the dynamics of banks’ stock prices, enriching the prospects for research in terms of methodological tools.

Nowadays we witness an increasing number of business frauds. To protect investors’ interest, a financial firm should possess an effective means to detect such frauds. In this regard, artificial neural networks (ANNs) are widely used for fraud detection. Traditional back-propagation-based algorithms used for training an ANN, however, exhibit the local optima problem, thus reducing the effectiveness of an ANN in detecting frauds. To alleviate the problem, this paper proposes an approach to training an ANN using an ant colony optimization technique, through which the local optima problem can be solved and the effectiveness of an ANN in fraud detection can be improved. Based on our approach, an associated prototype system is designed and implemented, and an exploratory study is performed. The results of the study are encouraging, showing the viability of our proposed approach.

We perform Markov chain Monte Carlo simulations for a Bayesian inference of the GJR-GARCH model which is one of asymmetric GARCH models. The adaptive construction scheme is used for the construction of the proposal density in the Metropolis-Hastings algorithm and the parameters of the proposal density are determined adaptively by using the data sampled by the Markov chain Monte Carlo simulation. We study the performance of the scheme with the artificial GJR-GARCH data. We find that the adaptive construction scheme samples GJR-GARCH parameters effectively and conclude that the Metropolis-Hastings algorithm with the adaptive construction scheme is an efficient method to the Bayesian inference of the GJR-GARCH model.

In this paper, a robot visual servo control algorithm is proposed by combining the conventional image based robot visual servoing algorithm with a trajectory estimation algorithm of a moving object using Kalman filter. The erroneous image information of a moving object due to the imprecise camera characteristics is compensated by applying Kalman filter to the process model of a moving object. The robot visual servo control algorithm is simulated, implemented and discussed with a Samsung FARA AT-2 robot and a MV50 Camera for its effectiveness, in both cases of with/without a trajectory estimation algorithm of a moving object using Kalman filter.

The paper entitled ”Quantum Quasi-Cyclic Low-Density Parity-Check Codes”, on pages 18-27 of this volume, has been retracted, because a large portion of the contents had been taken from the paper “Quantum Quasi-Cyclic LDPC Codes” by Manabu Hagiwara and Hideki Imai, published by arXiv.org in the year 2007.