Advanced Intelligent Computing Theories and Applications. With Aspects of Artificial Intelligence

In this paper, an effective artificial bee colony (ABC) algorithm is proposed to solve the multi-objective flexible job-shop scheduling problem with the criteria to minimize the maximum completion time, the total workload of machines and the workload of the critical machine simultaneously. By using the effective decoding scheme, hybrid initialization strategy, crossover and mutation operators for machine assignment and operation sequence, local search based on critical path and population updating strategy, the exploration and exploitation abilities of ABC algorithm are stressed and well balanced. Simulation results based on some widely used benchmark instances and comparisons with some existing algorithms demonstrate the effectiveness of the proposed ABC algorithm.

In this paper, an effective estimation of distribution algorithm (EDA) is proposed to solve the flexible job-shop scheduling problem with the criterion to minimize the maximum completion time (makespan). With the framework of the EDA, the probability model is built with the superior population and the new individuals are generated based on probability model. In addition, an updating mechanism of the probability model is proposed and a local search strategy based on critical path is designed to enhance the exploitation ability. Finally, numerical simulation is carried out based on the benchmark instances, and the comparisons with some existing algorithms demonstrate the effectiveness of the proposed algorithm.

The Inver-over operator holds a good result for small size Traveling Salesman Problem (TSP) while has worse capability for the large scale TSP. In this study, a Modified Inver-over operator is proposed to solve the TSP. In the Modified Inver-over operator, the direction of the tour is considered when applying the inversion and the city

c

is decided whether it is kept same after the inversion according to adaptively increasing probability, meanwhile, the

α

-nearest candidate set is used when selecting city

c

′

. We evaluate the proposed operator based on standard TSP test problems selected from TSPLIB and show that the proposed operator performs better than the Basic Inver-over operator and other operator in terms of solution quality and computational effort.

In this paper, a novel hybrid multi-objective particle swarm algorithm Mopsocd_BL is proposed to solve the flow shop scheduling problem with two objectives of minimizing makespan and the total idle time of machines. This algorithm bases on Baldwinian learning mechanism to improve local search ability of particle swarm optimization, and uses the Pareto dominance and crowding distance to update the solutions. Experimental results show that this algorithm can maintain the diversity of solutions and find more uniformly distributed Pareto optimal solutions.

This paper presents a discrete harmony search(DHS) algorithm for solving an n-job,m-machine lot-streaming flowshop scheduling problem(LFSP) with equal-size sublots, the objective is to minimize the total flow time. In the proposed DHS algorithm, The discrete job permutation without any conversion is used in the proposed DHS algorithm. To search for the best sequence,the DHS algorithm uses an effective initialization approach and a novel improvisation strategy. At the same time, an effective local search is embedded .Computational results demonstrate that the proposed DHS algorithm is very effectiveness for the lot-streaming flowshop scheduling problem.

Flow-shop scheduling problem (FSP) has been widely investigated in the area of manufacturing systems. Up to now, the NEH algorithm is the best heuristic approach to solve FSP. However, in large-scale problems, it takes quite long time for the NEH algorithm to find an approximate optimal solution. In this paper, two new techniques are proposed to improve the NEH algorithm. Firstly, to reduce the running time, block properties are developed and introduced to NEH algorithm. Secondly, to obtain solutions with smaller makespan, tie-break rules are applied. Simulation results show that these two techniques perform well in improving the NEH algorithm.

This paper addresses the flexible job shop scheduling problem with minimization of the makespan, maximum machine workload, and total machine workload as the objectives. A multiobjective memetic algorithm is proposed. It belongs to the integrated approach, which deals with the routing and sequencing sub-problems together. Dominance-based and aggregation-based fitness assignment methods are used in the parts of genetic algorithm and local search, respectively. The local search procedure follows the framework of variable neighborhood descent algorithm. The proposed algorithm is compared with three benchmark algorithms using fifteen classic problem instances. Its performance is better in terms of the number and quality of the obtained solutions.

In this study, we propose a genetic algorithm (GA) for the economic lot scheduling problem (ELSP) under extended basic period (EBP) approach and power-of-two (PoT) policy. The proposed GA employs a multi-chromosome solution representation to encode PoT multipliers and the production positions separately. Both feasible and infeasible solutions are maintained in the population through the use of some sophisticated constraint handling methods. Furthermore, a variable neighborhood search (VNS) algorithm is also fused into GA to further enhance the solution quality. The experimental results show that the proposed GA is very competitive to the best performing algorithms from the existing literature under the EBP and PoT policy.

In this paper, a multi-objective discrete harmony search algorithm (MDHS) is proposed to slove the lot-streaming flow shop scheduling problem with respect to the two objectives of makespan and total flow time. In the MDHS algorithm, the harmonies are represented as discrete job permutations, and an efficient initialization scheme, which is based on the famous NEH heuristic, is presented to construct the an initial solution in harmony memory. In addition, a local search approach based on insertion operator is embedded to improve the efficiency of the MDHS algorithm. Through the analysis of computational results, the proposed algorithm is superior to NEH heuristic algorithm.

Stochastic nature of vessel arrivals and handling times adds to the complexity of the well-known NP-hard berth allocation problem. To aid real decision-making under customer differentiations, a dynamic stochastic model designed to reflect different levels of vessel priorities is put forward. For exponential interarrival and handling times, a recursive procedure to calculate the objective function value is proposed. To reveal the characteristics of the model, numerical experiments based on heuristic approaches are conducted. Solution procedures based on artificial bee colony and genetic algorithms, covering both global and local search features, are launched to improve the solution quality. The practical inferences led by these approaches are shown to be helpful for container terminals faced with multifaceted priority considerations.

In this paper, we present a variable iterated greedy (vIGP_DE) algorithm where its parameters (basically destruction size and cooling parameter for the simulated annealing type of acceptance criterion) are optimized by the differential evolution algorithm. A unique multi-chromosome solution representation is presented such that first chromosome represents the destruction size and cooling parameter of the iterated greedy algorithm while second chromosome is simply a permutation assigned to each individual in the population randomly. As an application area, we choose to solve the no-idle permutation flowshop scheduling problem with the total flowtime criterion. To the best of our knowledge, the no-idle permutation flowshop problem hasn’t yet been studied thought it’s a variant of the well-known permutation flowshop scheduling problem. The performance of the vIGP_DE algorithm is tested on the Taillard’s benchmark suite and compared to a very recent variable iterated greedy algorithm from the existing literature. The computational results show its highly competitive performance and ultimately, we provide the best known solutions for the total flowtime criterion for the Taillard’s benchmark suit.

This paper considers a discrete artificial bee colony (DABC) algorithm for the blocking flow shop (BFS) scheduling problem to minimize total flowtime. The DABC algorithm utilizes discrete job permutations to represent food sources and applies discrete operators to generate new food sources for the employed bees, onlookers and scouts. First, an initialization scheme based on MME (combination of MinMax and NEH) heuristic is presented to construct an initial population with a certain level of quality and diversity. Second, a local search based on the insertion neighborhood is applied to onlooker stage to improve the algorithm’s local exploitation ability. Third, a destruction-construction operator is employed to obtain solutions for the scout bees. Computational simulations and comparisons show that the proposed algorithm (DABC) is effective and efficient for the blocking flow shop scheduling problems with total flowtime criterion.

Statistical textons has shown its potential ability in texture image classification. The maximal response 8 (MR8) method extracts an 8-dimensional feature set from 38 filters. It is one of state-of-the-art rotation invariant texture classification methods. This method assumes that each local patch has a dominant orientation, thus it keeps the maximal response from six responses of different orientations in the same scale. To validate whether local dominant orientation is necessary for texture classification, in this paper, a complex texton, complex response 8 (CR8), is proposed. The average and standard deviation of filter responses for different orientations is computed, and then an 8-dimensional complex texton is extracted. After using k-means clustering algorithm to learn a texton dictionary, a histogram of texton distribution could be built for a given image. Experimental results on one large public database show that CR8 could get comparable results with MR8.

In this paper, a perceptually motivated morphological strategy (PMMS) has been proposed to enhance the retrieval performance of common shape matching methods. We introduce a human perception custom that should be considered in a shape retrieval approach, and the proposed strategy based on the closing operation could simulate this custom properly. On the most widely used MPEG-7 dataset, we apply the proposed PMMS to improve the retrieval results of a popular shape matching method named Inner-Distance Shape Contexts (IDSC), and then we use the Locally Constrained Diffusion Process (LCDP) to further enhance the performance. This combination achieves a retrieval rate of 98.53%, which is the state-of-the-art performance on MPEG-7 dataset.

LBP operator is one of the best performing local texture descriptors and it has been broadly used in texture classification, face recognition, face expression recognition and so on. Existing improvements and applications of LBP are studied and summarized in this paper. Traditional LBP is reviewed first. Then several typical improvements are presented according to their different applications. The conclusion and possible future work are also suggested.

In this paper a new hybrid method has been proposed for solving the suspension design problem. A two-dimensional model of a car with linear passive suspension system and with two passengers has been considered. The vibration, experienced by the passengers due to road bump during vehicle motion has been minimized in time domain, by applying the proposed method. Moreover, the suspension parameters have been determined which satisfy performance as per ISO standards. The solutions/ parametric values so obtained have been further compared with the existing suspension parameters.

The present model is concerned with a multispecies fishery with two prey species both of which obeys the logistic law of growth and one predator species whose density is low. The predator consumes one of the prey species more intensively than the other because of its availability. For the predator species the growth function is taken as the model described by Smith. We assume that both the prey species are subjected to harvesting while the predator species is excluded from harvesting due to its low density. In this model we consider that the harvesting effort E is a function of time t. For optimization, we use the concept of generalized Legendre condition. The trajectory for the optimal singular extremal is derived and the optimal singular control is determined. Lastly, one numerical example is taken up and graph for steady state is drawn to illustrate the results.

Problem decomposition is the first step to apply a cooperative coevolutionary algorithm (CCEA) to a problem. This step determines how to divide the problem into components with an appropriate granularity. Most of the current methods implement a natural-based decomposition where each component plays a specific role or represents an emergent property. However, there could exist some real problems that the roles or the properties are hard to determine or somewhat unclear. This paper offers a solution by decomposing the problems in an unnatural way, which implements a blind decomposition. Our primary analysis indicates that the blind decomposition is feasible. We also provide some basic advice on how to implement the blind decomposition in combination with different collaboration methods.

General speaking, the heteroassociative morphological memory (HMM) is incomplete, namely, it cannot give a guarantee of perfect recall memory, even though without any input noises. The paper focuses on the problem and proposes a new method to improve performance of heteroassociative morphological memories. This method can realize the perfect recall of HMMs for perfect inputs or within a certain range of noises. An example is provided to illustrate the proposed method and its performance.

In this paper, a new approach of the restrained optimal perturbation method is firstly proposed to study the inverse problem in the reverse process of the one-dimensional convection diffusion equation, the idea of this method is brand new that in search for the optimal perturbation value by the given initial estimate, for determining the initial distribution based on the overspecified data, and the initial estimates plus optimal perturbation value can be treated as the final initial distribution, in order to overcome the ill-posedness of this problem, a regularization term is introduced in the objective functional. Numerical examples will be given, and the results show that our method is effective.

The blind separation of overdetermined mixtures, i.e., the case where more sensors than sources are available is considered in this paper. The contrast function for overdetermined blind source separation problem is presented, together with its gradient. An iterative method is proposed to solve the overdetermined blind source separation problem, where Gaussian mixture model is used to estimate the density of the unknown sources. The result of simulation demonstrates the efficiency of the proposed algorithm.

We present a new public-key encryption scheme, and prove its adaptive chosen-ciphertext security under the gap hashed Diffie-Hellman assumption in the standard model. Compared with previous public key encryption schemes with adaptive chosen-ciphertext security, our proposed scheme simultaneously enjoys the following advantages: small public key size, short ciphertext, low computational cost, weak complexity assumption and public verifiability.

In this paper we have proposed a lazy learning mechanism for multiclass classification using genetic programming. This method is an improvement of traditional binary decomposition method for multiclass classification. We train classifiers for individual classes for a certain number of generations. Individual trained classifiers for each class are combined in a single chromosome. A population of such chromosomes is created and evolved further. This method suppresses the conflicting situations common in binary decomposition method. The proposed lazy learning method has performed better than traditional binary decomposition method over five benchmark datasets taken from UCI ML repository.

Compared with value-function-based reinforcement learning (RL) methods, policy gradient reinforcement learning methods have better convergence, but large variance of policy gradient estimation influences the learning performance. In order to improve the convergence speed of policy gradient RL methods and the precision of gradient estimation, a kind of Actor-Critic (AC) learning algorithm based on incremental least-squares temporal difference with eligibility trace (iLSTD(

λ

)) is proposed by making use of the characteristics of AC framework, function approximator and iLSTD(

λ

) algorithm. The Critic estimates the value-function according to the iLSTD(

λ

) algorithm, and the Actor updates the policy parameter based on a regular gradient. Simulation results concerning a grid world with 10×10 size illustrate that the AC algorithm based on iLSTD(

λ

) not only has quick convergence speed but also has good gradient estimation.

K nearest neighbor algorithm (k-NN) is an instance-based lazy classifier that does not need to delineate the entire boundaries between classes. Thus some classification tasks that constantly need a training procedure may favor k-NN if high efficiency is needed. However, k-NN is prone to be affected by the underlying data distribution. In this paper, we define a new neighborhood relationship, called passive nearest neighbors, which is deemed to be able to counteract with the variation of data densities. Based on which we develop a new classifier called active and passive nearest neighbor algorithm (APNNA). The classifier is evaluated by 10-fold cross-validation on 10 randomly chosen benchmark datasets. The experimental results show that APNNA performs better than other classifiers on some datasets and worse on some other datasets, indicating that APNNA is a good complement to the current state-of-the-art of classification.

In many information card recognition systems, the most important task is to detect the correct location of the full-filling block. And it always needs high quality card and device. In this paper, a new noble recognition algorithm by support vector machines for user defined sheet made by normal paper is developed. We focused on recognizing the full-filling block in multi-noises environment. And we also focused on recognizing the sheet which has user defined format. The algorithm was also shown to be more effective and more robust than traditional recognition algorithm.

In this paper, a new multi-swarm method is proposed for multi-objective particle swarm optimization. To enhance the Pareto front searching ability of PSO, the particles are divided into many swarms. Several swarms are dynamically searching the objective space around some points of the Pareto front set. The rest of particles are searching the space keeping away from the Pareto front to improve the global search ability. Simulation results and comparisons with existing Multi-objective Particle Swarm Optimization methods demonstrate that the proposed method effectively enhances the search efficiency and improves the search quality.

This paper shows an application of Type-reduction algorithms for computing the steady state of an Interval Type-2 Fuzzy Markov Chain (IT2FM). The IT2FM approach is an extension of the scope of a Type-1 fuzzy markov chain (T1FM) that allows to embed several Type-1 fuzzy sets (T1FS) inside its

Footprint of Uncertainty

. In this way, a finite state Fuzzy Markov Chain process is defined on an Interval Type-2 Fuzzy environment, finding their limiting properties and its Type-reduced behavior. To do so, two examples are provided.

Reinforcement Learning, also sometimes called learning by rewards and punishments is the problem faced by an agent that must learn behavior through trial-and-error interactions with a dynamic environment [1]. With repeated trials however, it is expected that the agent learns to perfect its behavior overtime. In this paper we simulate the reinforcement learning process of a mobile agent on a grid space and examine the situation in which multiple reinforcement learning agents can be used to speed up the learning process by sharing their Q-values. We propose a sharing method which takes into consideration the weight of the experience acquired by each agent on the occasion of visiting a state and taking an action.

A novel variant of AdaBoost named AcoBoost is proposed to directly solve the asymmetric constraint optimization problem for cascade face detector using a two-stage feature selection approach. In the first stage, many candidate features are picked out by minimizing the weighted error. In the second stage, the optimal feature is singled out by minimizing the asymmetric constraint error. By doing so, the convergence rate is greatly speeded up. Besides, a new sample set called selection set is added into AcoBoost to prevent overfitting on the training set, which ensures good enough generalization ability for AcoBoost. The experimental results on building several upright frontal cascade face detectors show that the AcoBoost based classifiers have much better convergence ability and slightly worse generalization ability than the AdaBoost based ones. Some AcoBoost based cascade face detectors have satisfactory performance on the CMU+MIT upright frontal face test set.

In this paper, we present a translation model which uses syntactic structure and morphology of Myanmar language to improve Myanmar to English machine translation system. This system is implemented as a subsystem of Myanmar to English translation system and based on statistical approach by using Myanmar-English Bilingual corpus. It also uses two types of information: language model and translation model. The source language model is based on N-gram method to extract phrases from segmented Myanmar sentences and the translation model is based on syntactic structure, morphology of Myanmar language and Bayes rule to reformulate the translation probability. Experimental results showed that the proposed system gets a BLEU-score improvement of more than 22.08% in comparison with baseline SMT system.

We address the optimal design of a Distribution Network (DN), presenting a procedure employing Multi-Objective Genetic Algorithms (MOGA) to select the (sub) optimal DN configuration. Using multi-objective genetic optimization allows solving a nonlinear design problem with piecewise constant contributions in addition to linear ones. The MOGA application allows finding a Pareto frontier of (sub) optimal solutions, which is compared with the frontier obtained solving the same problem with linear programming, where piecewise constant contributions are linearly approximated. The two curves represent, respectively, the upper and the lower limit of the region including the real Pareto curve. Both the genetic optimization model and the linear programming are applied under structural constraints to a case study describing the DN of an Italian enterprise.

Buildings contribute a significant part in the energy consumption and CO2 emission in many countries. Building energy saving has thus become a hot research topic recently. The technology advances in power co-generation, on-site generation, and storage devices bring us the opportunity to reduce the cost and CO2 emission while meeting the demand in buildings. A fundamental difficulty to schedule this multi-energy system, besides other difficulties, is the discrete and large search space. In this paper, the multi-energy scheduling problem is modeled as a nonlinear programming problem with integer variables. A method is developed to solve this problem in two steps, which uses ordinal optimization to address the discrete and large search space and uses linear programming to solve the remaining sub-problems. The performance of this method is theoretically quantified, and compared with enumeration and a priority-and-rule-based scheduling policy. Numerical results show that our method provides a good tradeoff between the solution quality and the computational time comparing with the other two methods. We hope this work brings more insight on multi-energy scheduling problem in general.

In the paper, digital city is studied first, and then the smart traffic system is proposed. After that, a high-efficiency three levels intelligent incident detection algorithm is designed in detail. A better efficiency analysis by detecting rate, false alarm rate, and average detecting time, is obtained by simulation and experiment on Zhengzhou Ring Highway and BRT system. This algorithm not only fits flat plain, but also ring roads, therefore it can help city planning administrator implement the maximizing of traffic flows.

In this paper, a Genetic-AIS (Artificial Immune System) algorithm is introduced for PID (Proportional-Integral-Derivative) controller tuning using a multi-objective optimization framework. This hybrid Genetic-AIS technique is faster and accurate compared to each individual Genetic or AIS approach. The auto-tuned PID algorithm is then fused in an Immune feedback law based on a nonlinear proportional gain to realize a new PID controller. Immune algorithm presents a promising scheme due to its interesting features such as diversity, distributed computation, adaptation and self monitoring. Accordingly, this leads to a more effective Immune-based tuning than the conventional PID tuning schemes benefiting a multi-objective optimization prospective. Integration of Genetic-AIS algorithm with Immune feedback mechanism results into a robust PID controller which is ultimately evaluated via simulation control test scenarios to demonstrate quick response, good robustness, and satisfactory overshoot and disturbance rejection characteristics.

As the popularity of Web 2.0 increases, web users used to establish communication relationships and share resources with others online. Recently, one important research for retrieving information from the web has become a major issue. Moreover, due to the social network popularity, there is much more opportunity to find out various human resources to help people get what they want. In our work, we propose to help the social network users to find out the most reachable experts who are authorized researchers on a common topic area with the users. For this purpose, we propose how to calculate centrality closeness from requesting user to experts, analyze the relevancy of an expert to the topic, and finally combine both of the result to rank the most reachable experts for the requesting user.

The Forex intervention is a main instrument to restrain the unexpected float of exchange rate. This paper constructs a simultaneous equations model (SEM) with the central bank reaction function and return of exchange rate to address the effects of Forex intervention. We find that the lagged intervention is a valid and rational instrumental variable to correct this simultaneity bias of Forex intervention and volatility. We also find that BoJ Forex intervention is more effective in “leaning with the wind” than in “leaning against the wind” condition, but unavoidably increases the volatility in either case, and consequently brings more financial risk.

The performance of the wireless communication system depends on the designer’s understanding to the regular pattern of environment electrical noise distribution in which the communication system is located. The lack of the knowledge of the underground coal mine’s electrical noise distribution usually led to the current common practice as directly adoption the terrestrial wireless communication systems property index when design the underground coal mine’s wireless communication system. The huge difference does exist between the underground coal mine’s electrical noise distribution and the terrestrial counterpart, which usually causes the underground wireless communication system’s decline in the communication quality, or not functioning properly. As the special circumstances of underground coal mine makes it difficult to apply the terrestrial electrical noise measurement method in the underground, this article presents a new method for measuring the electrical noise in underground coal mine, which has been successfully applied in measuring the typical underground coal mine’s electrical noise distribution.

In this paper, a current controller for grid-connected inverter is proposed by using a fuzzy logic control algorithm. In PI controller to control the grid-connected inverter, the gains of PI controller are changed with the aid of the fuzzy logic algorithm in order to get the fast transient performance despite of the input variations and load disturbances. The inputs of fuzzy logic controller are the error between the measured currents and the reference values in rotating reference frame, and the derivation of regulated voltage. The effectiveness of proposed controller strategy has been verified by simulation with PSIM software and compared with that of the conventional PI controller.

In static Wireless sensor networks position, at present using a mobile beacon assisted positioning mechanism. However, in assisted positioning of the mobile beacon, the mobile path of beacon nodes has a great influence to positioning performance, the existing mobile assisted positioning algorithm consider less to path planning, for its lack, paper proposes a more flexible dynamic heuristic path planning method, with more practical for the application of irregular network topology.

IEC61850 communication standard for digital substation automation creates a new way to think about conventional protection scheme and configuration of substation. The presence of communication link in process bus makes a revolutionary change for future digital substation. This paper discusses briefly about process bus architecture and presents a new approach of thinking in busbar structure to make the bus protection less sensitive with communication link performance degradation. IEC61850-9-2LE implementation is analyzed and process bus based on Ethernet technology is considered. Transient element based conventional bus differential protection is used with split busbar configuration and finally demonstrates the reliability and feasibility of proposed technique for digital substation.

This paper proposes a new sliding mode observer (SMO) based anti-windup PI speed control strategy for Permanent Magnet Synchronous Motor (PMSM). A SMO is constructed to estimate the back-electromotive-force (EMF) signals robustly, and then the position and speed of a PMSM can be calculated according to the back-EMF equations. Based on the estimated position and speed of PMSM, four different anti-windup PI strategies are used and compared for speed control of PMSM, The simulation results show the superior control performance compared with conventional schemes.

This paper studies principles, characteristics and process of psychological warfare and the essentials of artificial intelligence. By providing the theoretical frame of expert system in psychological warfare (ESPW) based on production rule, this paper makes breakthroughs on the combination of artificial intelligence and psychological warfare. This theoretical frame is the foundation of ESPW and it covers production rule set, fact database, knowledge reason tree, reason machine principle and other contents.

In the field of rotating machinery diagnosis using traditional intelligent diagnosis method, the state judgment and fault detection are usually carried out by symptom parameters (SPs). However, it is difficult to find the general and highly sensitive SPs for rotating machinery diagnosis. Intelligent methods, such as neural networks, genetic algorithms, etc., often cannot converge when being trained. In order to solve these problems, this paper proposes a new intelligent diagnosis based on distinctive frequency components (DFCs) and support vector machines (SVMs) which can be used to detect faults and recognize fault types of rotating machinery. The method has been applied to detect the structural faults of rotating machinery, and the efficiency of the method is verified by practical examples.

With the increasing application fields of domain ontology models, how to develop the ontology model for a specific domain has become an urgent problem to solve. This paper presents and summaries three comparatively mature ontology engineering methodologies proposed respectively by Uschold & King, Grüninger & Fox and Methontology against the standard of IEEE1074-1995. Then the paper analyses the shortcomings of the three methodologies in the processes of meta-data model development and makes the proposition that the converse development method for domain ontology model is a beneficial supplement for traditional positive development method.

Today RDF data are proliferating so fast that RDF query engines are faced with very large graphs that contain thousand million of RDF triples. Often there are a lot of joins should be processed when using RDF query language-SPARQL execute queries and the key issue for optimizing SPARQL execution plans is join ordering so selectivity estimation is very important to query cost. Exact estimation could optimize query and reduce query time, in contrast bad estimation could misguide the order of joins and increase query cost. In this paper we introduce two selectivity estimation methods: method based on histogram and method based on Index. We analyze the execute details of each method and compare the two methods then we give the conclusion that which method are better when execute query in large dataset. Finally, our experimental (using these two methods in different queries that have different join times in different size datasets.) testify our viewpoint.

This paper presents two user interfaces for a pattern-based User Interactive Question Answering system, which are designed to encourage users to ask questions by using semantic patterns. One is a Guide-Based User Interface (GBUI), which can guide users with clear instructions. However, it involves many steps and the operation may become tedious. The other is a Recommendation-Based User Interface (RBUI), which recommends a few relevant patterns containing automatically suggested details for each free-text question. However, the recommended patterns may not always be satisfactory and sometimes the user’s revision is needed. In comparing these two user interfaces, we propose a new Complexity Evaluation Model (CEM) to evaluate the complexity on the basis of user log study and a realistic focused user study. The results of the study user logs, which cover a test set of 1605 users and 488 semantic patterns, show that RBUI can improve the complexity of GBUI by 39.8% on average. The improvement is also confirmed by the user study. It has thus become clear that RBUI can improve the usability of the UIQA system in terms of helping the system accumulate high quality pattern-based questions.

Wireless Sensor Networks (WSNs) coverage optimization is to maximize the coverage of WSNs while keeping service quality. This paper extends the problem to 2.5D and studies PSO based WSNs coverage optimization on Digital Elevation Models (DEMs). To compute network coverage on DEMs, a method of computing individual sensor node coverage is introduced. This paper also proposes an improved algorithm based on Dissipative Particle Swarm Optimization (DPSO). Simulation experiments show that the algorithm can effectively improve WSNs coverage on DEMs.

This paper proposes a real-time algorithmic framework for Automatic Speech Recognition (ASR) in presence of multiple sources in reverberated environment. The addressed real-life acoustic scenario definitely asks for a robust signal processing solution to reduce the impact of source mixing and reverberation on ASR performances. Here the authors show how the implemented approach allows to improve recognition accuracies under real-time processing constraints and overlapping distant-talking speakers. A suitable database has been generated on purpose, by adapting an existing large vocabulary continuous speech recognition (LVCSR) corpus to deal with the acoustic conditions under study.

Network security situation assessment is the core of situation awareness, and is also a qualitative and quantitative description of network security state. In this paper, the security state and intrusion alarm event in the network or the host system have corresponded to the state and observation symbols in HMM, so that a network security situation evaluation model based on HMM has been proposed. This model intrudes the alarm sequences generated from the detection system through association analysis to calculate the risk index of each host, so as to give a quantitative evaluation of the risk status of the whole net-work system. The Network Risk Index can be calculated easily and quickly. The experimental results show that this model can effectively and accurately give a quantitative evaluation of the security state of the network system.

In this paper we propose a novel Rearranged Modular 2DPCA (Rm2DPCA) algorithm for face recognition. In the proposed algorithm, the original images are first divided into modular images. Then, the sub-images are rearranged to form a 2D matrix. A covariance matrix is constructed directly using all the arranged matrices, and its eigenvectors are derived for image feature extraction. Experiments compared with other similar approaches show that this method has two advantages: One is its better recognition performance, the other is less computational cost.

Pose and illuminations remain great challenges to current face recognition technique. In this paper, visible image (VI) and near-infrared image (NIR) are fused for performance improvement. When directional binary code is adopted as feature representation, AdaBoost algorithm and the cascade structure are used for classification. Fusion is done at decision level and classification scores are normalized using three different rules, i.e. Min-Max, Z-Score and Tanh-Estimators. Experimental results suggest that the proposed algorithm using VI achieve better performance than NIR when pose and expression variations are present. However, NIR shows much better robustness against illumination and time difference than VI. Due to the complementary information available in two image modalities, fusion of NIR and VI further improves the system performance.

The fingerprint image quality is a key factor on the match results since it will cause spurious and missed minutiae when matching with the low quality images. It is important to estimate the image quality to guide the feature extraction and matching. In this paper we investigate the specifications that can reflect the image quality such as orientation coherence, core position and so on. We define a quasi core as a stable point to examine the validity of the captured position. We apply the idea of penalty function in the optimization theory to combine the specifications to get a quality score. The method is robust since it investigates the quality specifications entirely. The testing results on FVC database are given to verify the feasibility and effectiveness.

In this paper, we propose a novel texture classification method based on feature extraction through c-means clustering on the contourlet domain. In particular, all the features representing each contourlet subband are extracted by a c-means clustering standard algorithm. By investigating these features, we use the weighted

L

1

-norm for comparing the features of the two corresponding subbands of two images and define a new distance between two images. According to the new distance, a

k

-Nearest Neighbor (kNN) classifier is utilized to perform texture classification (TC), and experimental results reveal that our proposed approach outperforms two current state-of-the-art texture classification approaches.

In iris recognition systems how to represent texture pattern is an important issue. The paper proposes a novel approach based on SIFT for feature representation of iris texture. This approach partitions a normalized iris image into non-overlapping small sub-images and uses SIFT descriptor for representing the characteristics of each sub-image. As such the iris texture pattern is represented by an ordered-set of SIFT descriptors. This representation is very distinctive and insensitive to illumination changes. In addition, it encodes the positional information of iris texture pattern. For iris matching we use Bhattacharyya distance to measure the dissimilarity between two SIFT descriptors. The final distance is a sum of the distances of the corresponding pairs of SIFT descriptors in two iris images. The experimental results on UBIRIS.v1 and UBIRIS.v2 show that proposed method has promising performance.

Correct wood recognition has an important meaning in rational use of wood resources. Automatic wood recognition based on wood stereogram are studied in this paper. According to the wood stereogram characteristics, a method of image normalization is presented firstly. Then wood texture features are extracted using Gabor wavelet with analyzing the best scale and orientation parameters. In addition to the mean and standard deviation on the Gabor filter bank, entropy, contrast and other statistical features are used for classification. Experimental results show that the entropy can better extract texture features on Gabor wavelet, which greatly improve the wood recognition rate.

Automatic age estimation based on facial images is important but challenging in face recognition research. A Super-Resolution Reconstruction algorithm was proposed to implement the age estimation of facial images, which cut the facial image into small pieces. Then after building high resolution images by using Super-Resolution Reconstruction algorithm, the RBF neural networks was used to train and test these high resolution images. At last, the classifier ensemble with genetic algorithm was used to estimating age information. Finally, experimental results demonstrate that it is an effective method.

This paper presents a wearable physical activity sensor system and its activity classification algorithm. In addition, we investigate possible combinations of different sensor placements, and identify an optimal combination to achieve the best classification performance. The sensor system consists of several sensor modules that can be synchronized to record the accelerations of diverse motions/activities. In our experiment, three sensor modules are mounted on participants’ hand wrists, waists, and ankles, respectively, to collect seven categories of activity accelerations. The proposed classification algorithm consisting of acceleration acquisition, signal preprocessing, feature generation, and feature reduction, is capable of translating time-series acceleration signals into important time- and frequency-domain feature vectors. The dimension of features is reduced by linear discriminate analysis (LDA), and then the reduced features are sent to a

k

-nearest neighbor (

k

-NN) classifier for classification. Our experimental results have successfully validated the effectiveness of the proposed classification algorithm. The best classification accuracy is 96.98% when the sensor modules are placed on hand and ankle simultaneously.

This paper implements Biomimetic Pattern Recognition (BPR) into the adaptive classification of motor imagery EEG data by introducing three adaptive operators. The adaptive BPR works well when facing the problems of non-stationary in on-line BCI. We have conducted some experiments in adaptive scheme and the results demonstrate that our new algorithm is efficiency and robust compared with non-adaptive classifiers.

Fingerprinting is used to determine originators of unauthorized copies. Multiple users may collude by creating an average or median of their individual fingerprinted copies. Early fingerprinting research including ACC (anti-collusion code) cannot support large number of users. There have been two fingerprint researches for practically large user group support: SACC (scalable ACC) scheme and MC-CDMA (multi-carrier code-division multi-access) based fingerprinting scheme. In SACC scheme, they use a codebook extending ACC using Gaussian distributed random variable and use angular decoding scheme for average, median, LCCA attack robustness. MC-CDMA scheme uses direct spreading approach for identifying large group of users. In this paper, we compare two schemes in three aspects: imaging quality, computational complexity, and user capacity. Our experimental results show that SACC scheme has achieved better imaging quality and lower computational complexity which are important for video fingerprinting performance. MC-CDMA scheme outperforms SACC in user capacity as MC-CDMA uses direct spreading based CDMA approach while SACC uses a frequency hopping based CDMA approach.

This paper introduces an algorithm to super-resolve an image based on a self-training filter (STF). As in other methods, we first increase the resolution by interpolation. The interpolated image has higher resolution, but is blurry because of the interpolation. Then, unlike other methods, we simply filter this interpolated image to recover some missing high frequency details by STF. The input image is first downsized at the same ratio used in super-resolution, then upsized. The super-resolution filters are obtained by minimizing the mean square error between the upsized image and the input image at different levels of the image pyramid. The best STF is chosen as the one with minimal error in the training phase. We have shown that STF is more effective than a generic unsharp mask filter. By combining interpolation and filtering, we achieved competitive results when compared to support vector regression methods and the kernel regression method.

As fire accidents usually cause economic and environmental damage, including the loss of human lives, video-based fire detection has become more appealing in surveillance systems. However, video based fire detection algorithms demand tremendous computational and I/O requirements. To meet these requirements, we introduce an SIMD (Single Instruction Multiple Data) based multi-core architecture that consists of 16 processing elements (PEs) and small local memory. In addition, we compare the performance and efficiency of the multi-core architecture with a commercial Texas Instrument digital signal processor (TI DSP) to demonstrate the potential for improved performance of the multi-core architecture. Experimental results indicate that the multi-core architecture is 27.18 times and 3.89 times better than TI DSP in terms of execution time and energy efficiency, respectively.

Plant classification is very important and necessary with respect to agricultural informization, ecological protection and plant automatic classification system. In this paper, we present a multilinear independent component analysis (MICA) algorithm and apply it to a multimodal plant leaf recognition problem involving multiple leaves imaged in different periods and illuminations. To show the validity of the method, we apply it to a plant leaf image dataset. The experimental results show that the method is efficient and feasible.

In this paper we have adapted an agent approach for traffic light control. According to this, the proposed system contains agents and their world which in turn contains roads, cars, traffic lights etc. Each of these agents observe the traffic density and control the traffic light at the junction by using observe-think-act rule i.e. the agents will continuously observe the traffic and depending on the density and waiting time it decides which rule can be inferred and finally it implements the condition to the traffic light controller which can efficiently manage the traffic flow near the junction. The system has been implemented by using NetLogo based traffic simulator. The investigation is to control the traffic at the road junction by applying few inference rules.It reduces the waiting time and increases the efficiency of the traffic light controller intelligently.

In this paper, a new mass segmentation algorithm is proposed. In the new proposed algorithm, a fully automatic marker-controlled watershed transform is first proposed to segment the mass region roughly, and then a level set is used to refine the segmentation. The new algorithm combines the advantages of both methods. The combination of the watershed based segmentation and level set method can improve the efficiency of the segmentation. Images from DDSM were used in the experiments and the results show that the new algorithm can improve the accuracy of mass segmentation.

Texture segmentation is a preliminary step in a wide spectrum of computer vision applications. Although the search for robust texture descriptors has been going for decades, there is still lack of texture features providing theoretical as well as practical evidence for successful segmentation. In this paper a novel algorithm for estimating scale exponents is described and applied in texture segmentation task. The estimated scale exponents are linearly dependent on generalized fractal dimensions. It has been proven that fractal dimensions are invariant under bi-Lipshitz transforms [1], which are general smooth transforms including perspective transforms. We estimate scale exponents in blocks around feature points, which allow us to characterize local regions and further segment them. In the case of finite resolution images, the proposed estimation algorithm produces robust to rotations and illumination changes features as predicted theoretically. The extracted features are applied for unsupervised segmentation using c-means fuzzy clustering by estimating spatial membership functions for each texture cluster. We experimented with textures from well-known Brodatz[2] and Vistex[3] texture database.

In this paper, an improved prototyping method is adopted to perform the task of ageing a human face, which aims to incorporate sparseness constrained NMF to extract texture features of facial image and find out which part of the factorized matrix should be kept sparse. The experimental results show that NMF with coefficient H sparse is more capable of feature extraction compared to PCA method in the course of texture aging.

This paper introduces a robotic wheelchair that can automatically move alongside a caregiver. Because wheelchair users are often accompanied by caregivers, it is vital to consider how to reduce a caregiver’s load and support their activities, while simultaneously facilitating communication between the caregiver and the wheelchair user. Moreover, a sociologist pointed out that when a wheelchair user is accompanied by a companion, the latter is inevitably seen by others as a caregiver rather than a friend. In other words, the equality of the relationship is publicly undermined when the wheelchair is pushed by a companion. Hence, we propose a robotic wheelchair able to move alongside a companion, and facilitate easy communication between the companion and the wheelchair user. Laser range sensors are used for tracking the caregiver and observing the environment around the wheelchair. To confirm the effectiveness of the wheelchair in real-world situations, we conducted experiments at an elderly care center in Japan. Results and analyses are also reported in this paper.

This paper presents a novel approach to mobile robot exploration. In this approach, mobile robots send their local maps to the central controller and coordinate with each other using a wireless sensor network (WSN). Different from existing rendezvous point-based exploration strategies, the use of a WSN as the communication media allows quick and cost-effective exploration and mapping of an unknown environment. Overall, this paper introduces WSN-aided mobile robot exploration strategy and shows comparative performance evaluations using the Player/Stage simulation platform. Here, our main goal is to present potential advantages of WSN-aided mobile robot exploration for Simultaneous Localization and Mapping (SLAM).

Detecting moving objects from video frames is one of the key techniques in computer vision. Background subtraction is a common way to detect moving objects at present. A new background subtraction algorithm is proposed in this paper. The algorithm describes backgrounds by a combination of hue and improved local binary pattern (LBP) texture and adopts the idea of Gaussian mixture model that uses multiple modes to represent background. In order to reduce matching complexity and satisfy real-time, the LBP texture feature vectors are simplified. Experiments show that the proposed algorithm can satisfy real-time in common resolution videos, can remove effectively the effect of shadow and can detect moving objects more accurately than others.

The research presents a preliminary approach to perform any type of image processing task using 16-bit digital signal controllers. Even though this attempt is aimed at Edge Detection, the research opens up possibilities for numerous other algorithms of signal and image processing that can be implemented using the same low cost hardware. FPGA’s & DSP’s are widely used to perform hardware-based signal processing task. It is an efficient but generally an expensive solution for image processing applications. On the other hand a conventional 8-bit MCU doesn’t have enough capability to handle memory intensive DSP algorithms. In this regard, a digital signal controller offers a tradeoff between cost and performance.

In this paper, we propose an efficient video analysis framework to assign broadcast soccer video shots into their respective view classes, and then detect players in long view shots. Our technique is built on dominant color region based segmentation for soccer playfield extraction. A long-view shot classifier uses a combination of “grass-area” ratio and “top-grass” analysis. A player detector applies the distinctive uniform knowledge of interesting objects based on colors referring from the result of playfield. In order to verify the player region segmented using colour, we introduce the four-seed edge features which prune the redundant edges denoting the noise of court lines or audiences. The player detection performance is suitable to employ tracking methods in order to exploit higher semantic information from the games. Experimental evaluation of the framework is extensively demonstrated in numerous challenging test sequences of the 2010 FiFa World Cup South Africa. The results show the robustness of our framework, and the potential future-work.

This paper proposes a method for building detection and 3D reconstruction of building face from sparse view of monocular camera. According to this method, building faces are detected by using color, straight line, edge and vanishing point. In the next step, building faces from multi view are extracted. Point clouds of building face are obtained from triangulation step. The building faces are reconstructed by plane fitting afterward. The simulation results will demonstrate the effectiveness of this method.

Traffic video analysis is a challenging problem: crowded moving vehicles with various appearances, illumination changes, and speed variations according to the traffic flow. In this paper, we propose an efficient single-camera Traffic Monitoring System (TMS), which is capable of automatically analyzing the vehicles flow on urban streets in real time. The system has three main modules included calculating density of vehicles based on background subtraction methods, estimating average speed of traffic flow using optical flow method and counting the number of vehicles on the street by clustering motion features relied on Delaunay Triangulation algorithm. From that fundamental information, our system infers several high semantic events such as traffic jams, breaking-law vehicles, people crossing street. Experiments are demonstrated in real-life scenarios with heavy traffic in Ho Chi Minh City, Vietnam.

An online gesture recognition system using a dynamical programming, One-Pass DP, is proposed in this paper. Firstly, 8 directions of hand motions are extracted with skin color analysis and optical flow calculation using a primary visual cortex model. Then, the patterns of motion are used to compose 40 basic templates of gestures. At last, hand gestures are recognized by the One-Pass DP algorithm. Experiments dealt with individual and compound gestures were executed by online processing, the results confirmed the effectiveness of the proposed system.

Dynamic hand gesture tracking and recognition system can simplify the way humans interact with computers and many other non-critical consumer electronic equipments. This system is based on the well-known “Wave Controller” technology developed at the University of Wollongong [1–3] and certainly a step forward in video gaming and consumer electronics control interfaces. Many computer interfaces used today such as keyboard, mouse, joystick or gaming wheels have constrained the artistic ability of many users, as they are required to respond to the computer through pressing buttons or moving other apparatus. Most of the drawbacks of the modern interfaces can be tackled by using a reliable hand gesture tracking and recognition system based on both Lucas-Kanade and Moment Invariants approaches. The realtime functional ability of this system will enhance the user experience as users are no longer have any physical connection to the equipment being controlled.

Image quality assessment traditionally means the comparison of original image with its distorted version using conventional methods like

Mean Square Error

(MSE) or

Peak Signal to Noise Ratio

(PSNR). In case of Blind Quality Evaluation with no prior knowledge about the image, a single parameter becomes insufficient to define the overall image quality. This paper proposes a quality metric based on

sharpness

of the image, presence of

noise

, overall

contrast

and

luminance

of the image and the

detection of the eyes

. Experimental results reveal that the proposed metric has strong relevance with human quality perception.

This paper presents an efficient palmprint based recognition system. In this system, the image is divided into disjoint sub-images. For each sub-image, the dominant orientation pixels based on the force field transformation are identified. Structure tensor values of these dominant orientation pixels of each sub-image are averaged to form tensor matrix for the sub-image. Eigen decomposition of each tensor matrix is used to generate the feature matrix which is used to take decision on matching. The system has been tested on IITK database. The experimental results reveal the accuracy of 100% for the database.

This paper presents a modified geometric hashing technique to index the database of facial images. The technique makes use of minimum amount of search space and memory to provide best matches with high accuracy against a query image. Features are extracted using Speeded-Up Robust Features (SURF) operator. To make these features invariant to translation, rotation and scaling, a pre-processing technique consisting of mean centering, principal components, rotation and normalization has been proposed. The proposed geometric hashing is used to hash these features to index each facial image in the database. It has achieved more than 99% hit rate for top 4 best matches.

In this paper, the robust asymptotical stability is investigated for a class of interval neutral systems. Based on Lyapunov stable theory, the delay-dependent criteria are derived to ensure the global, robust, asymptotical stability of the addressed system. The criteria can be checked easily by LMI control toolbox in Matlab. A numeric example is given to illustrate the effectiveness and improvement over some existing results.

In this paper, the exponential stability of a class of nonlinear switched systems with time-varying disturbances is considered. A new piecewise time-varying Lyapunov functionals which are decreasing at switching times by construction are introduced to investigate exponential stability of switched delay systems, sufficient conditions expressed as linear matrix inequalities are obtained.

Masses are the typical signs of breast cancer. Correctly classifying mammographic masses as malignant or benign can assist radiologists to diagnosis breast cancer and can reduce the unnecessary biopsy without increasing false negatives. In this paper, we investigate the classification of masses with level set segmentation and shape analysis. Based on the initial contour guided by the radiologist, level set segmentation is used to deform the contour and achieve the final segmentation. Shape features are extracted from the boundaries of segmented regions. Linear discriminant analysis and support vector machine are investigated for classification. A dataset consists of 292 ROIs from DDSM mammogram images were used for experiments. The method based on Fourier descriptor of normalized accumulative angle achieved a high accuracy of Az=0.8803. The experimental results show that Fourier descriptor of normalized accumulative angle is an effective feature for the classification of masses in mammogram.

Isometric feature mapping (ISOMAP), locally linear embedding (LLE) and Laplacian eigenmaps (LE) are recently proposed nonlinear dimensionality reduction methods of manifolds. When these methods are satisfied with some specific constraints, some hidden connections can be found between principal component analysis (PCA) and those manifolds learning based approaches. In this paper, some derivations are presented to validate the idea and then some conclusions are drawn.

This paper puts forward a useful method for step length adaptation of the mutation distribution in ES- using the GPC (Generalized Predictive Control) to adapt the global step size. Similar to the concept of evolution path, the mutation step is the function of historical information generated by the iterative processes of ES algorithm. In our method, the ES algorithm is regarded as a controlled system and modeled as a CARIMA (Controlled Auto Regressive Integrated Moving Average) model. The parameters of CARIMA model are estimated by RLS (the recursive least squares) with forgetting factor, and then the current global optimum step size (the control parameter) is calculated by the GPC to feed back to ES, the output and the control quantum are used to estimate the parameters of CARIMA model iteratively.

A video shot segmentation scheme with dual-detection model is proposed. In the pre-detection round, the Uneven Blocked differences are presented and used in Adaptive Binary Search (ABS) to detect shot boundaries. In the re-detection round, the Scale Invariant Feature Transform (SIFT) method is applied to exclude false detections. Experiments show that this algorithm achieves well performances in detecting both abrupt and gradual boundaries.

Memristor behavior is explained with a physical model based on quantum mechanics that claims charge is naturally created anytime energy is absorbed at the nanoscale. Quantum mechanics requires specific heat to vanish at the nanoscale, and therefore the electrical resistive heating in the memristor cannot be conserved by an increase in temperature. Conservation proceeds by frequency up-conversion of the absorbed energy to produce photons that in submicron thin films have energy beyond the ultraviolet. By the photoelectric effect, the photons create excitons inside the memristor that decrease resistance only to be recovered later in the same cycle as the electrons and holes of the excitons are attracted to and destroyed by the polarity of the voltage terminals. Observed memristor behavior is therefore the consequence of excitons being created and destroyed every cycle.

Various studies on generating test data have been done up to date, but few test data generated by these studies can effectively detect faults lying in the program. We focus on the problem of generating test data for both paths coverage and faults detection. First, the problem above is formulated as a bi-objective optimization problem with one constraint, whose two objectives are the number of faults detected in the traversed path and the risk level of these faults, respectively, and the unique constraint is that the traversed path is just the target one; then, a multi-objective evolutionary algorithm is employed to effectively solve the formulated model; finally, the proposed method is applied in

bubble sort program

manually injected with some faults, and compared with the random method and the evolutionary optimization one without the task of detecting faults. The experimental results confirm the advantage of our method.

For the problem that a millimeter wave (MMW) image contains noise and behaves low resolution, a novel MMW image reconstruction method, combined the non-negative sparse coding shrinkage (NNSCS) technique and the partial differential equations (PDEs) algorithm (denoted by NNSCS+ PDEs), is proposed in this paper. The method of PDEs is an efficient image reconstruction technique and is easy to implement. However, MMW image is highly contaminated by much unknown noise, and the reconstruction result is not satisfied only using PDEs to process images. While the NNSCS only relies on the high-order statistical property of an image and is a self-adaptive image denoising method. Thus, combined the advantage of NNSCS and PDEs, the MMW image can be well restored. In test, a natural image is used to testify the validity of the NNSC+PDEs method, and the signal noise ratio (SNR) is used as the measure criterion of restored images. Compared with NNSCS and PDEs respectively, simulation results show that our method is indeed efficient in the task of reconstructing WWM images.

The embedded Ethernet becomes an important communication way of supporting the development of internet of things. Accordingly, the embedded Ethernet technology is divided into two types: “micro controller & Ethernet control chip” solutions and “single- chip” solutions. However, the former is the popular choice. Freescale’s highly-integrated 32-bit microcontroller MCF52259 is used as main controller for developing the embedded Ethernet, which bases on the V2 ColdFire micro-architecture. During the process of constructing, embedded component is used as the guidance. Firstly, the hardware circuit design of Ethernet drive is implemented, and then the process of realization of embedded Ethernet drive software is given. Finally, thorough verification is performed by compiled test cases. In this way, it can be proved that the design has excellent stability and encapsulation.

This paper proposes a novel image magnification method based on bilinear interpolation, wavelet, and partial differential equation (PDE) techniques. The image which is interpolated linearly is decomposed by wavelet into a low frequency component image and three high frequency component images, and then the three high frequency component images and the original image regarded as low-frequency component will be used for image magnification by invert wavelet transform. Finally, a PDE involving gray fidelity constraint item called improvement-self-snake mode is presented in post-processing of the magnified image. The experimental results show that the proposed linear interpolation-wavelet-PDE approach is indeed efficient and effective in image magnification. In addition, we also compare the signal-to-noise ratio (SNR) of the linear interpolation-wavelet-PDE magnification method with methods of linear interpolation, linear interpolation-wavelet, and wavelet-PDE. The simulating results show that the linear interpolation-wavelet-PDE method indeed outperforms the three kinds of image magnification approaches mentioned above.

Utilizing the variation in the detection sensitivity of the catalytic sensor under different temperatures, a new method of analyzing inflammable gases with a single catalytic sensor based on thermostatic detection and RBF neural network theory is proposed. A mathematical model of analyzing different inflammable gases is constructed based on dynamic learning algorithm. Experiments were carried out with sample mixed gases of firedamp, carbon monoxide and hydrogen. The results show that the mixed inflammable gases can be effectively analyzed by the single catalytic sensor.

A novel palmprint recognition method using the fast sparse coding (FSC) algorithm is proposed in this paper. This algorithm is based on iteratively solving two convex optimization problems, the

L

1

-regularized least squares problem and the

L

2

-constrained least squares problem. As the same as the standard sparse coding (SC) algorithm, this FSC algorithm can model the receptive fields of neurons in the visual cortex in brain of human, however, it has a faster convergence speed than the existing SC model. Using this FSC algorithm, feature basis vectors of palmprint images can be learned successfully. Here, the PolyU palmprint database, used widely in palmprint recognition research, is selected as the test database. Furthermore, utilizing learned palmprint features and the radial basis probabilistic neural network (RBPNN) classifier, the task of palmprint recognition can be implemented efficiently. Using the recognition rate as the measure criterion, and compared our palmprint recognition method with principal component analysis (PCA), standard SC and fast independent component analysis (FastICA), the simulation results show further that this method proposed by us is indeed efficient in application.

When using probabilistic neural network (PNN) to recognize human speaker, there exists structure complex problems if the training sample amount is large and the redundancy degree is high. To overcome this shortcoming, this paper proposes a method of principal component analysis (PCA) for keeping the effective information and reducing the redundancy of characteristic parameters, that means, this method can reduce the dimension of input data and optimize the structure of PNN network successfully. Experimental results show that the proposed speaker recognition method based on the combination of principal component analysis (PCA) and probabilistic neural network (PNN) is an effective and reliable new speaker recognition system.

In this study, a ranking test problem of Computer Adaptive Testing (CAT) is benchmarked by employing three popular classifiers: Artificial Neural Network (ANN), Support Vector Machines (SVMs), and Adaptive Network Based Fuzzy Inference System (ANFIS) in terms of ordinal classification performances. As the pilot test, “History of Civilization” class which offered in Bahcesehir University is selected. Item Response Theory (IRT) is focused for the determination of system inputs which are item responses of students, item difficulties of questions, and question levels. Item difficulties of questions are Gaussian normalized to make ordinal decisions. The distance between predicted and expected class values is employed for accuracy estimation. Comparison study is conducted to the ordinal class prediction correctness and performance analysis which is observed by Receiver Operating Characteristic (ROC) graphs. The results show that ANFIS has better performance and higher accuracy than ANN and SVMs in terms of ordinal question classification when the ordinal decisions are practically made by Gaussian Normal Distribution and ROC graphs are focused to observe any significant difference among the performances of classifiers.