Search

This paper focuses on traffic volume forecasting that is an essential component of any responsive traffic control or route guidance system. A new approach for traffic volume prediction is proposed based on wavelet transform and neural networks. First, apply multi-resolution analysis to the original traffic volume time series to obtain a trend series and a hierarchy of detail series. Then apply neural networks to each obtained time series. Next sum all the forecasting values to get the final prediction of traffic volume. This hybrid method is implemented within a Matlab environment. The feasibility of the developed method as well as its validity to predict traffic volume has been demonstrated on real data gathered in Suzhou city. Moreover, a comparison between the hybrid method and conventional neural networks is conducted. The results show the proposed hybrid model outperformed the neural networks. [ABSTRACT FROM AUTHOR]

In this paper, a radial basis function (RBF) network based method inspired by mature algorithms for face recognition is applied to classify traffic scenes in single image mode. Not to follow traditional ways of estimating traffic states through image segmentation and vehicle tracking, this method avoids complicated problems in digital image processing (DIP) and can operate on just one image, while the old ones rely on consecutive images. The proposed method adopts discrete cosine transform (DCT) for feature selection, then a supervised clustering algorithm is fulfilled to help design hidden layer of RBF network for which Gaussian function is chosen, finally linear least square (LLS) is used to solve the weights training problem. Experiments show that this method is valid and effective under the new application background. [ABSTRACT FROM AUTHOR]

Traffic signs can provide drivers with very valuable information about the road, in order to make driving safer and easier. In recent years, traffic signs recognition has aroused wide interests in many scholars. It has two main parts- the detection and the classification. This paper presents a new method for traffic signs classification based on probabilistic neural networks (PNN) and Tchebichef moment invariants. It has two hierarchies: the first hierarchy classifier can coarsely classify the input image into one of indicative signs, warning signs or prohibitive signs according to its background color threshold; the second hierarchy classifiers including of three PNN networks can concretely identify traffic sign. The inputs of every PNN use the new developed Tchebichef moment invariants. The simulation results show that the two-hierarchy classifier can improve the classification ability meanwhile can use in real-time system. [ABSTRACT FROM AUTHOR]

Robust and reliable vehicle detection is a challenging task under the conditions of variable size and distance, various weather and illumination, cluttered background, the relative motion between the host vehicle and background. In this paper we investigate real-time vehicle detection using machine vision for active safety in vehicle applications. The conventional search method of vehicle detection is a full search one using image pyramid,which processes the image patches in same way and costs same computing time, even for no vehicle region according to prior knowledge. Our vehicle detection approach includes two basic phases. In the hypothesis generation phase, we determine the Regions of Interest (ROI) in an image according to lane vanishing points; furthermore, near, middle, and far ROIs, each with a different resolution, are extracted from the image. From the analysis of horizontal and vertical edges in the image, vehicle hypothesis lists are generated for each ROI. Finally, a hypothesis list for the whole image is obtained by combining these three lists. In the hypothesis validation phase, we propose a vehicle validation approach using Support Vector Machine (SVM) and Gabor feature. The experimental results show that the average right detection rate reach 90% and the average execution time is 30ms using a Pentium(R)4 CPU 2.4GHz. [ABSTRACT FROM AUTHOR]

The automatic detection and recognition of road signs play important role in the driver assistance systems and can increase the safety on the roads. In this paper we propose a system of a road signs classifier which is based on ensemble of the non Euclidean distance neural networks and an arbitration unit. The input to this system comes from the sign detection module which supplies a normalized, binarized and resampled pictogram of a detected sign. The system performs classification on deformable models. The classifier is composed of a mixture of experts (binary distance neural networks) operating on slightly tilted or shifted versions of pictograms. This ensemble of experts is orchestrated by an arbitration module which operates in the winner-takes-all mode with a novel modification of promoting the most populated group of unanimous experts. The experimental results showed great robustness of the system and very fast response time which is an important factor in the driving assistance systems. [ABSTRACT FROM AUTHOR]

In this paper, a novel recognition system based on rough neural network is presented for the application of vehicle classification in wireless sensor network. The proposed system is evaluated using real-world signal datasets as well as two conventional methods. Compared with them, approach based on rough neural network achieves high performance improvement. Furthermore, the purposed system is extended for multi-channel sensor data fusion directly. Since the experiment results are attractive, an algorithm based on rough neural network is believed to have potential for applications of recognition and data fusion in wireless sensor networks. [ABSTRACT FROM AUTHOR]

The number of vehicles is rapidly increased as modern industrialization is developed worldwide. Vehicle recognition has been studied for some time because many people acknowledged it is critical information to solve vehicle-related problems. However, few researchers have tried to recognize the model of vehicles. In this paper a novel method is proposed to recognize vehicles' model corresponding manufacturers in order to increase the efficiency of recognition. Texture descriptors are computed from the front image of vehicles. Then, a three-layer neural network was built and trained with the texture descriptors for model recognition. The proposed method demonstrates 94% recognition rate for moving vehicles' models. [ABSTRACT FROM AUTHOR]

This paper presents an intelligent vehicle security system for handling the vehicle thefts problem under the framework of capturing and analyzing dynamic human behaviors. Since human driving skill is a kind of dynamic biometrical feature which is complex and difficult to imitate, it is unique and more secure than static features such as password, fingerprint and face. By utilizing this dynamic property we focus on the research ideal of classifying the drivers into authorized ones and unauthorized ones by modeling their individual driving performance. Firstly, we develop an experimental system architecture. We collect the data of steering, acceleration and braking directly from human driving behaviors as inputs to the system, which aims to achieve better robustness and efficiency. Then, we use fast fourier transform (FFT), principal component analysis (PCA) and independent component analysis (ICA) for data reduction. The features extracted are sent to support vector machine (SVM) for learning and recognition. In the next step, we embed the intelligent classifier into a security system to identify the authorized drivers in response to the real time driving performances. Finally, the experimental results verify that the proposed method is valid and useful against the vehicle thefts problem with a success rate of around 80%. [ABSTRACT FROM AUTHOR]

In this paper, we propose robust adaptive neural network (NN) control for helicopter systems by using the Implicit Function Theorem and the Mean Value Theorem, which are useful tools for handling nonlinear nonaffine systems. We focus on single-input single-output (SISO) helicopter systems, which are exemplified by certain single-channel modes of operation, such as vertical flight and pitch regulation, and also by special conditions under which the multiple channels become decoupled. It is shown that under the proposed NN control, the output tracking error converges to a small neighbourhood of the origin, while all closed loop signals are Semi-Globally Uniformly Ultimately Bounded (SGUUB). [ABSTRACT FROM AUTHOR]

This paper models the ambient noise as α stable distribution and proposes a novel robust multiuser detection (MUD) method that involves an adaptive nonlinear preprocessor based on multilayer perceptron neural-network whose action is to suppress the negative effect of impulsive noises on the followed decorrelating decision-feedback (DDF) multiuser detector. Simulation results indicate the proposed new method is robust and offers performance enhancement over traditional technology in impulsive noises. [ABSTRACT FROM AUTHOR]

This paper studies the performance of extreme learning machine with complex-valued radial basis function (ELM-CRBF) in the channel equalization applications. Comparing with complex minimal resource allocation network (CMRAN), complex radial basis function (CRBF) network and Bayesian equalizers, the simulation results show that ELM-CRBF equalizer is superior in terms of symbol error rate (SER) and learning speed. [ABSTRACT FROM AUTHOR]

Recently, a new learning algorithm for the feedforward neural network named the complex extreme learning machine (C-ELM) which can give better performance than traditional tuning-based learning methods for feedforward neural networks in terms of generalization and learning speed has been proposed by Huang et al. In this paper, we propose a new widely linear recursive C-ELM algorithm for nonlinear channel equalizer. The proposed algorithm improves its performance especially in case of real valued modulation such as BPSK and PAM. The computer simulation results demonstrate the improvement in performance achievable with the proposed equalization algorithm. [ABSTRACT FROM AUTHOR]

Mobile communications and digital wireless communications are requested high frequency use-rates, more efficient data transmission with limited signal power, frequency band. As multiple users share the same frequency in the mobile communications environment, the spectrum efficiency is getting higher. Moreover, as the effect of the velocity of the mobile object and the terrain surroundings get higher, the digital modulation method is required that the character of linear constant amplitude. In this paper, to restore simply and correctly the received signal of quadrature phase shift keying (QPSK) signal in digital wireless communications, we suggest and simulate an algorithm for detection of QPSK signal using time delay neural networks (TDNN). As the results of simulation, the suggested method is confirmed that the phase information of the QPSK signal is recovered simply and correctly in the mobile communications and digital wireless communications. [ABSTRACT FROM AUTHOR]

This paper presents an efficient carrier frequency offset (CFO) estimation algorithm for the orthogonal frequency-division multiplexing (OFDM)-based wireless local area networks (WLANs) which adopts the IEEE 802.11 standardization. This algorithm adopts the BP network with the orthogonal projector for the fast adaptive CFO estimation. Numerical results are presented to demonstrate the effectiveness of the proposed algorithm. [ABSTRACT FROM AUTHOR]

The recommendation G.728 depends on the Levinson-Durbin (L-D) algorithm to update its gain filter coefficients. In this paper, it is contrasted with BP neural network method. Because quantizer has not existed at optimizing gain filter, the quantization SNR can not be used to evaluate its performance. This paper proposes a scheme to estimate SNR so that gain predictor can be separately optimized with quantizer. Using BP neural network filter, the calculation quantity is only 6.7 percent of L-D method's and its average segment SNR is about 0.156dB higher than G.728. It is also used to evaluate the case that excitation vector is 16 or 20 samples, respectively, the BP neural network algorithm has similarly good result. [ABSTRACT FROM AUTHOR]

Application layer multicast routing problem is a multi- objective optimization problem. This paper reviews the multicast routing problem, analyzes the multi-constraints, especially the importance of network layer load balance. Then a modified SOFM approach is proposed to build a multicast tree according to cost, application layer load balance and network layer load balance. Numerical simulations show that compared with existent routing algorithms, the proposed algorithm improved the performance on all the three metrics. [ABSTRACT FROM AUTHOR]

This paper addresses the important issues of error control for video transmission over 3G, considering the fact that wireless video delivery faces the huge challenge of the high error rate and time variability in wireless channel. This paper proposes a real world statistics based event-trigger bit error rate (BER) model, which can describe and handle the time-varying wireless channel error characteristics better. Moreover, a recurrent neural network is employed to decide the state transfer as a mechanism. Simulation results and comparisons demonstrate effectiveness and efficiency of the proposed method in term of visual performance and transmission efficiency over a variety of wireless channel conditions. [ABSTRACT FROM AUTHOR]

A novel multilevel quantifying spread spectrum PN sequence based on the chaos of Cellular Neural Network (CNN) is proposed in this paper. The chaotic sequences are created from a CNN with three cells and multilevel quantified to be the PN sequences for the spread spectrum communication systems (SSCS). And then better sequences are filtered out in terms of the equilibria points, self-correlation and mutual-correlation of the chaotic PN sequences. These PN sequences can provide more enhanced multiple access capacity and robustness to the noises than conventional m- and Gold sequences because of the sensitivity to the initial conditions of the chaotic sequences and the good dynamical performance of the CNN. The filter processing helps the SSCS to resist on the rake declination and the interference from other users. The experiment results show that the proposed chaotic PN sequences are much better than the conventional sequences for SSCS. [ABSTRACT FROM AUTHOR]

A new hyper-chaos communication system based on cellular neural network (CNN) is proposed in this paper. Hyper-chaos is generated with fifteen-cell CNN, then it's transferred to a digital sequence. The chaotic sequence is better than gold sequence when they are utilized in direct sequence spread spectrum system. Compared with the traditional gold sequence system, there is 3dB improvement in CNN chaotic sequence system when the channel is multi-path channel. Because of the complex dynamic behavior of hyper-chaos, security signal could be transferred through wireless channel. The structure of hardware CNN spread spectrum system is also shown, and the security of CNN communication is analyzed at last. [ABSTRACT FROM AUTHOR]

By applying tomography theory which is highly developed in fieldssuchas medical computerized tomography and seismic tomography to communication network, network tomography has become one of the focused new technologies, which can infer the internal performance of the network by external end-to-end measurement. In this paper, we propose a novel Inference algorithm based on the recurrent multilayer perceptron (RMLP) network capable of tracking nonstationary network behavior and estimating time-varying, internal delay characteristics. Simulation experiments demonstrate the performance of the RMLP network. [ABSTRACT FROM AUTHOR]

A new wavelet-based neural network architecture employing the BiLinear Recurrent Neural Network (BLRNN) for time-series prediction is proposed in this paper. It is called the Multiscale BiLinear Recurrent Neural Network (M-BLRNN). The wavelet transform is employed to decompose the time-series to a multiresolution representation while the BLRNN model is used to predict a signal at each level of resolution. The proposed M-BLRNN algorithm is applied to the long-term prediction of network traffic. The performance of the proposed M-BLRNN algorithm is evaluated and compared with the traditional MultiLayer Perceptron Type Neural Network (MLPNN) and the BLRNN. The results show that the M-BLRNN gives a 20.8% to 76.5% reduction in terms of the normalized mean square error (NMSE) over the MLPNN and the BLRNN. [ABSTRACT FROM AUTHOR]

In this paper, queue length in the buffer is considered as a scale to evaluate the quality of service of communication network. To realize that the system is stable and avoid the data congestion under complex environments, the neural networks predictor and controller are designed, which can predict the bursty available bandwidth for ABR traffic effectively and force the queue level in the buffer to the desired region, respectively. The fairness of different connections is achieved through fair algorithm. The simulations verify the effectiveness of the proposed scheme. [ABSTRACT FROM AUTHOR]

This paper presents a secure scheme to enhance the transmission security of content-based hidden biometric data over insecure network. Encryp-tion, coding, spread spectrum modulation, and data hiding techniques are used to improve the security and secrecy of the transmitted templates. Secret keys are generated by the biometric image and used as the parameter value and initial condition of the chaotic maps, and each transaction session has different secret keys to protect from the attacks. Templates are encrypted by chaotic encryption, encoded by BCH encoding method, modulated by chaotic neural network parameter modulation (CNNPM), and then hid into the cover image. Experi-mental results show that the security, performance, and accuracy of the presented scheme is encouraging comparable with other methods found in the current literature. [ABSTRACT FROM AUTHOR]

This paper proposes a new approach to build lightweight Intrusion Detection System (IDS) based on Random Forest (RF). RF is a special kind of ensemble learning techniques and it turns out to perform very well compared to other classification algorithms such as Support Vector Machines (SVM) and Artificial Neural Networks (ANN). In addition, RF produces a measure of importance of feature variables. Our approach is able not only to show high detection rates but also to figure out stable output of important features simultaneously. The results of experiments on KDD 1999 intrusion detection dataset indicate the feasibility of our approach. [ABSTRACT FROM AUTHOR]

A new network intrusion detection system is presented in this paper. The system is skillfully combined with fuzzy technique and neural network which architecture and arithmetic is redesigned. In order to overcome the difficulty of specifying the membership function of rules depending on experiences of experts in multi-dimensional space, fuzzy neural network model is introduced to carry through proper nonlinear division of input/output characteristics of complex system and to generate fuzzy rule sets and added membership relation automatically. The new system architecture adopts the network processor to collect and analyze the data in the low layer of network, and a prototype system is established. This prototype system behaves better ability of intrusion detection and lower rate of distort, and that it has the ability to detect unknown attack. [ABSTRACT FROM AUTHOR]

This paper proposes a method to detect network intrusions by using the PCASOM (principal components analysis and self-organizing map) neural networks. A modified unsupervised learning algorithm which is more suitable for intrusion detection is presented. Experiments are carried out to illustrate the performance of the proposed method by using DARPA 1998 evaluation data sets. It shows that the proposed method can cluster the network connections into proper clusters with high detection rate and relatively low false alarm rate. [ABSTRACT FROM AUTHOR]

The Intrusion Detection System (IDS) is generally used the misuse detection model based on rules because this model has low false alarm rates. However, the rule based IDSs are not efficient for mutated attacks, because they need additional rules for the variations of the attacks. In this paper, we propose an intrusion detection system using the Principal Component Analysis (PCA) and the Time Delay Neural Network (TDNN). Packets on the network can be considered as gray images of which pixels represent bytes of the packets. From these continuous packet images, we extract principal components. And these components are used as an input of a TDNN classifier that discriminates between normal and abnormal packet flows. The system deals well with various mutated attacks, as well as well-known attacks. [ABSTRACT FROM AUTHOR]

In this paper, the Self Organizing Maps (SOM) learning and classification algorithms are firstly modified. Then via the introduction of match-degree, reduction-rate and quantification error of reducing sample, a novel approach to intrusion detection based on Multi-layered modified SOM neural network and Principal Component Analysis (PCA) is proposed. In this model, PCA is applied to feature selection, and Multi-layered SOM is designed to subdivide the imprecise clustering in single-layered SOM layer by layer. Experimental results demonstrate that this model can provide a precise and efficient way for implementing the classifier in intrusion detection. [ABSTRACT FROM AUTHOR]

Pattern recognition and learning of unknown worms have become a bottleneck of network security since a lot of variants of old worms and new worms occurred. To overcome this bottleneck, many traditional approaches were tested but failed. In this paper, a normal model of a web system was proposed to detect all selfs and all non-selfs, especially all unknown worms. The normal model was built on the 2-dimension attributes of space and time of the system. Moreover, a BP neural network was used to design an adaptive learning mechanism of the immunized web system. The non-self learning was utilized to recognize most unknown worms through the trained BP network, which was trained with the feature data in the worm database. Besides, the innate non-self selection was designed to recognize all known worms. Experiments validated effectiveness of this approach on the BP network and the normal model. [ABSTRACT FROM AUTHOR]

In this paper, a new and efficient image hiding scheme is proposed. Different from the existing methods, the secret data is embedded into the prediction errors produced by the neural network nonlinear predictor, and the non-uniform quantization method is used to embed secret data. The proposed method can achieve higher embedding payload while keeping smaller distortion, and experimental results are given to show the advantage of this scheme. [ABSTRACT FROM AUTHOR]

Information hiding capacity of digital image is the maximum information that can be hidden in an image. But the lower limit of information hiding, the minimum detectable information capacity is also an interesting problem. This paper proposes a method to analyze the minimum detectable capacity of information hiding in digital images based on the theories of attractors and attraction basin of neural network. The results of research show that the attractors of neural network decide the lower limit of information hiding. [ABSTRACT FROM AUTHOR]

This paper proposes a new watermarking scheme in which a logo watermark is embedded into the spatial domain of the original image using Back-Propagation neural networks (BPN). BPN will learn the characteristic of the image, and then watermark is embedded and extracted by the trained BPN. The image is divided into 8(8 blocks and the average pixel value of each block is used as the desired output value of the BPN. The quantized DC coefficient of discrete cosine transform (DCT) domain of each block is used as input value of the BPN to be trained. After the BPN is trained using those input/output values, watermark is embedded into the spatial domain using the trained BPN. The trained BPN also used in watermark extracting process. Experimental results show that the proposed method has good imperceptibility and high robustness to common image processing. [ABSTRACT FROM AUTHOR]

In this paper, we propose a novel invisible watermarking method with image signature, which can resist not only common image attacks but also copy attack. The image signature algorithm referred is based on Image Independent Block Feature (IIBF) obtained by Independent Component Analysis (ICA). Simulations demonstrate the robustness of this method. In addition, this invisible watermarking method can be applied to protect the visible-watermarked image. That is to say, we can claim the ownership of the image after visible watermark removal by the invisible watermark. [ABSTRACT FROM AUTHOR]

In this paper, a novel robust halftone image watermarking algorithm based on neural networks is presented. By using neural networks, the error diffusion filter can be adjusted adaptively, which can integrate the data embedding with the halftone processing in our scheme. Experimental results show that the proposed halftone image watermarking is invisible and robust against various signal processions such as JBIG compression, noise adding, cropping, daubing and print-scan. Especially, the algorithm can extract the watermark without the help from the original image. [ABSTRACT FROM AUTHOR]

A multi-bit digital audio watermarking scheme based on blind source separation (BSS) is proposed in this paper. The host audio signal is first permuted and divided into frames equally, and the watermark embedding and extraction procedure are executed frame by frame. In embedding procedure, the embedding watermark signal is generated firstly, and then mixed together with the permuted host audio signal through a BSS model. After that, the obtained mixtures are combined and inverse permuted to form the watermarked audio signal. In extraction procedure, watermarked audio signal is permuted and split into two signals, and then BSS technique is applied to obtain the separated embedding watermark signal, by observing the correlation values between it and the embedding watermark signal, the existence of watermark can be determined. With the secret keys, we can recover the watermark image further more. Experimental results show the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Temporal synchronization is a problem that cannot be ignored in most video watermarking application. In this paper, a video watermarking scheme is proposed to resist temporal desynchronization by using shot segmentation and two-dimensional principal component analysis (2DPCA). In this scheme, the to-be-watermarked video is first segmented into different shots. And each shot is analyzed by 2DPCA in xt tomography to get its principal feature frames (PFFs). And then the watermark is embedded into the PFFs. This scheme is robust to temporal desynchronization through a content dependent way. Simulations show the feasibility and reliability of this scheme. It is also robust to most of the common video processing/attacks, such as MPEG2 compression, filtering, frame resizing, noising, etc. [ABSTRACT FROM AUTHOR]

The image encryption based on blind source separation (BSS) takes advantage of the underdetermined BSS problem to encrypt multiple confidential images. Its security can be further improved if the number of images to be simultaneously encrypted increases. However, the BSS decryption speed will correspondingly decrease since the computational load of the BSS algorithms usually has nonlinear relation with the number of the source signals. To solve the problem, this paper presents a fast decryption algorithm based on adaptive noise cancellation by using the knowledge of the key images, which are used in the BSS-based method and available at the receiving side. As a result, the number of the source signals for the fast BSS decryption is decreased in half, and the decryption time is considerably reduced. Both computer simulations and performance analyses demonstrate the efficiency of the proposed method. [ABSTRACT FROM AUTHOR]

How to design an efficient and security keyed hash function is always the point in modern cryptography researches. In this paper, A better chaos sequence is generated by RBF neural network through training the known chaotic sequence generated by a piecewise nonlinear map, then the sequence is used to construct keyed hash function. One advantage of the algorithm is that the hidden-mapping model of neural network makes it difficult to get the direct mapping function of the ordinary chaos hash algorithm. Simulation results show that the keyed hash function based on the neural network has good one-way, weak collision, better security property and it can be realized easily. [ABSTRACT FROM AUTHOR]

In self and non-self discrimination (SNSD) model, it is very important to generate a desirable detector set since it decides the performance and scale of the SNSD model based task. By using the famous principle of negative selection in natural immune system, a novel generating algorithm of detector, multiple-point bit mutation method, is proposed in this paper. It utilizes random multiple-point mutation to look for non-self detectors in a large range in the whole space of detectors, such that we can obtain a required detector set in a reasonable computation time. This paper describes the work procedure of the proposed detector generating algorithm. We tested the algorithm by using many datasets and compared it with the Exhaustive Detector Generating Algorithm in details. The experimental results show that the proposed algorithm outperforms the Exhaustive Detector Generating Algorithm both in computational complexities and detection performance. [ABSTRACT FROM AUTHOR]

The characteristics of erotic images are analyzed in this paper and a novel algorithm for erotic images recognition is proposed. The algorithm first obtains the mask images of the recognized image by skin color detecting and texture analyzing, and then locates the possible position of human trunk in mask image according to the established model of trunk, based on which the characteristics of erotic images are extracted. Furthermore, the SVM classifier is used to recognize the erotic images based on those extracted characteristics. The experimental results show that the recognition accuracy rate of the proposed algorithm is higher than other algorithms and the proposed algorithm is efficient and effective. [ABSTRACT FROM AUTHOR]

In this paper, we present a unified framework for sensor validation, which is an extremely important module in the engine health management system. Our approach consists of several key ideas. First, we applied nonlinear minor component analysis (NLMCA) to capture the analytical redundancy between sensors. The obtained NLMCA model is data driven, does not require faulty data, and only utilizes sensor measurements during normal operations. Second, practical fault detection and isolation indices based on Squared Weighted Residuals (SWR) are employed to detect and classify the sensor failures. The SWR yields more accurate and robust detection and isolation results as compared to the conventional Squared Prediction Error (SPE). Third, an accurate fault size estimation method based on reverse scanning of the residuals is proposed. Extensive simulations based on a nonlinear prototype non-augmented turbofan engine model have been performed to validate the excellent performance of our approach. [ABSTRACT FROM AUTHOR]

Maximizing the productivity in semiconductor manufacturing, early detection of process and/or equipment abnormality. Since most of the key processes in semiconductor production are performed under extremely high vacuum condition, no other action can be taken unless the undergoing process is terminated. In this paper, time series based neural networks have been employed to assist the decision for determining potential process fault in real-time. Principal component analysis (PCA) for the dimensionality reduction of the data is first performed to handle smoothly in real-time environment. According to the PCA, 11 system parameters were selected, and each of them were then classified using modeled and tested in time series. Successful detection on different types of process shift (or fault) was achieved with 0% false alarm. [ABSTRACT FROM AUTHOR]

After finishing the design of the sound reproduction system, it is necessary and important to evaluate how much each component of the sound reproduction system affects the final quality of audio. In this paper, we present an intelligent diagnostics for sound reproduction system. It is based on PEAQ(ITU-R BS. 1387) and artificial neural networks. This method makes it possible to consider the human preferences for various genres of music which differ from nation to nation and culture to culture. [ABSTRACT FROM AUTHOR]

While conventional approaches to diagnostics focus on detecting and identifying situations or behaviors which have previously been known to occur or can be anticipated, anomaly detection focuses on detecting and quantifying deviations away from learned "normal" behavior. A new anomaly detection scheme based on Growing Structure Multiple Model System(GSMMS) is utilized in this paper to detect and quantify the effects of slowly evolving anomalies on the crankshaft dynamics in a internal combustion engine. The Voronoi sets defined by the reference vectors of the growing Self-Organizing Networks(SONs), on which the GSMMS is based, naturally form a partition of the system operation space. Regionalization of system operation space using SONs makes it possible to model the system dynamics locally using simple models. In addition, the residual errors can be analyzed locally to accommodate unequally distributed residual errors in different regions. [ABSTRACT FROM AUTHOR]

This paper applies two ANN methods-RBF and SOFM on fault diagnosis for two-shaft aero-engines. Two-shaft aero-engines are complex rotating machines which have many components and high rotating speed. First we presented both the principles and advantages of RBF and SOFM neural networks. Second we described the fundamentals of two-shaft aero-engines vibration fault diagnosis, and then obtained the standard fault samples (training samples) and simulation samples (testing samples). Third we applied the two ANN methods to perform diagnosing. The accurate diagnostic results have proved the effectiveness of the RBF and SOFM methods for vibration fault diagnosis of two-shaft aero-engines. Finally, the relative advantages and disadvantages of the two ANN methods are contrasted, and suggestions can be obtained on when one might use one of the two methods. [ABSTRACT FROM AUTHOR]

Operating the nuclear power generations safely is not easy way because nuclear power generations are very complicated systems. In the main control room of the nuclear power generations, about 4000 numbers of alarms and monitoring devices are equipped to handle the signals corresponding to operating equipments. Thus, operators have to deal with massive information and to analyze the situation immediately. In this paper, the fault diagnosis system is designed using 2-steps neural networks. This diagnosis method is based on the pattern of the principal variables which could represent the type and severity of faults. [ABSTRACT FROM AUTHOR]

In this paper, the neural network is employed to learn the complicated association of the stock information stream with stock price movement. In general, the positive stream of stock information is assumed to stimulate buying and increase the stock prices, and the negative stream of stock information is assumed to result in selling and decrease the stock prices. Based on the stock information entropy, the intensity for stock information stream is measured and applied to associating with the stock price movement with the aid of neural networks. Experiments illustrate the associations statistically. The results are helpful in probing the microstructure of the stock markets from a new angle. [ABSTRACT FROM AUTHOR]

This paper discusses the application of support vector machine (SVM) in stock price change trend forecasting. By reviewing prior research, thirteen technical indicators are defined as the input attributes of SVM. By training this model, we can forecast if the stock price would rise the next day. In order to make best use of market information, analyst recommendations about upgrading stocks are employed. So we put forward an improved method to evaluate if an upgrade classification of SVM is reliable. In our method, recommendation accuracy is first calculated according to historical advice. Then the more objective relative accuracy is deduced by considering the influence of total stock market index. Moreover, improved model is examined with the real data in Shanghai stock exchange market. Finally, we discuss some interesting hints to help readers understand this model more explicitly. [ABSTRACT FROM AUTHOR]

This paper presents a novel method for predicting the stock index based on the multiclass classification. The strategy firstly discretizes the stock index to different interval and assigns a class label to each interval, which yields a multiclass classification problem. After briefly reviewing multiclass classification algorithm, a multiclass classifier based on analytical center of version space is proposed and applied to stock index prediction. Experiments on shanghai stock index demonstrates that the strategy of stock index prediction proposed is validated and of practical value. [ABSTRACT FROM AUTHOR]

This paper investigates whether it is possible to exploit the nonlinear behavior of daily returns to improve forecasting on Chinese Shanghai stock market index over short and long horizons. We compare out-of-sample forecasts of daily returns for the Chinese Shanghai Stock Market Index, generated by five competing models, namely a linear AR model, the LSTAR and ESTAR smooth transition autoregressive models and two ANN models: MLP and JCN. The research results show that the nonlinear ANN models may be an appropriate way to improve forecasts. The return on the Chinese Shanghai Stock Market Index could be predicted more accurately by using ANN models, and the neural network technique could be said to represent a slight improvement in prediction of the stock index with respect to AR model and STAR models. [ABSTRACT FROM AUTHOR]