Neural Computing and Applications

Cho, Siu-Yeung; Wong, Jia-Jun

doi: 10.1007/s00521-007-0108-8pmid: N/A

This paper describes a novel method of facial representation and recognition based upon adaptive processing of tree structures. Instead of the conventional flat vector representation for a face, a neural network approach-based technique is proposed to transform the Localised Gabor Feature (LGF) vectors extracted from human facial components into Human Face Tree Structure (HFTS) to represent a human face. A structural training algorithm is assigned to train and recognize the face identity in this HFTS representation with the corresponding LGF vectors. By benchmarking using the tested public face databases presented in this paper, our approach is able to achieve accuracy up to 90% under different scenarios of lighting conditions and posture orientations.

Mohamed Ben Ali, Yamina

doi: 10.1007/s00521-007-0114-xpmid: N/A

Training neural networks is a complex task provided that many algorithms are combined to find best solutions to the classification problem. In this work, we point out the evolutionary computing to minimize a neural configuration. For this purpose, a distribution estimation framework is performed to select relevant features, which lead to classification accuracy with a lower complexity in computational time. Primarily, a pruning strategy-based score function is applied to decide the network relevance in the genetic population. Since the complexity of the network (connections, weights, and biases) is most important, the cooling state of the system will strongly relate to the entropy as a minimization function to reach the desired solution. Also, the framework proposes coevolution learning (with discrete and continuous representations) to improve the behavior of the evolutionary neural learning. The results obtained after simulations show that the proposed work is a promising way to extend its usability to other classes of neural networks.

Akin, Mehmet; Kurt, Muhammed; Sezgin, Necmettin; Bayram, Muhittin

doi: 10.1007/s00521-007-0117-7pmid: N/A

We developed a new method for estimation of vigilance level by using both EEG and EMG signals recorded during transition from wakefulness to sleep. Previous studies used only EEG signals for estimating the vigilance levels. In this study, it was aimed to estimate vigilance level by using both EEG and EMG signals for increasing the accuracy of the estimation rate. In our work, EEG and EMG signals were obtained from 30 subjects. In data preparation stage, EEG signals were separated to its subbands using wavelet transform for efficient discrimination, and chin EMG was used to verify and eliminate the movement artifacts. The changes in EEG and EMG were diagnosed while transition from wakefulness to sleep by using developed artificial neural network (ANN). Training and testing data sets consist of the subbanded components of EEG and power density of EMG signals were applied to the ANN for training and testing the system which gives three situations for the vigilance level of the subject: awake, drowsy, and sleep. The accuracy of estimation was about 98–99% while the accuracy of the previous study, which uses only EEG, was 95–96%.

Wei, Xun-Kai; Li, Ying-Hong; Li, Yu-Fei; Zhang, Dong-Fang

doi: 10.1007/s00521-007-0113-ypmid: N/A

A novel machine learning paradigm, i.e., enclosing machine learning based on regular geometric shapes was proposed. First, it adopted regular minimum volume enclosing and bounding geometric shapes (sphere, ellipsoid, box) or their unions and so on to obtain one class description model. Second, Data description, two class classification, learning algorithms based on the one class description model were presented. The most obvious feature was that enclosing machine learning emphasized one class description and learning. To illustrate the concepts and algorithms, a minimum volume enclosing ellipsoid (MVEE) case for enclosing machine learning was then investigated in detail. Implementation algorithms for enclosing machine learning based on MVEE were presented. Subsequently, we validate the performances of MVEE learners using real world datasets. For novelty detection, a benchmark ball bearing dataset is adopted. For pattern classification, a benchmark iris dataset is investigated. The performance results show that our proposed method is comparable even better than Support Vector Machines (SVMs) in the datasets studied.

Chtourou, S.; Chtourou, M.; Hammami, O.

doi: 10.1007/s00521-007-0116-8pmid: N/A

Noisy and large data sets are extremely difficult to handle and especially to predict. Time series prediction is a problem, which is frequently addressed by researchers in many engineering fields. This paper presents a hybrid approach to handle a large and noisy data set. In fact, a Self Organizing Map (SOM), combined with multiple recurrent neural networks (RNN) has been trained to predict the components of noisy and large data set. The SOM has been developed to construct incrementally a set of clusters. Each cluster has been represented by a subset of data used to train a recurrent neural network. The back propagation through time has been deployed to train the set of recurrent neural networks. To show the performances of the proposed approach, a problem of instruction addresses prefetching has been treated.

Ji, Luping; Yi, Zhang; Shang, Lifeng

doi: 10.1007/s00521-007-0119-5pmid: N/A

To develop new image processing applications for pulse coupled neural network (PCNN), this paper proposes an improved PCNN model by redesigning the linking input, activity strength, linking weight, pulse threshold and pixel update rule. Two typical image processing examples based on such a model, namely fingerprint orientation field estimation and noise removal, are presented for explaining how to use the PCNN and determine parameters in image processing. Experiments show that the improved model is quite useful, and the PCNN-based approaches achieve better image processing results than the traditional ones.

Nguwi, Yok-Yen; Kouzani, Abbas

doi: 10.1007/s00521-007-0120-zpmid: N/A

An automatic road sign recognition system first locates road signs within images captured by an imaging sensor on-board of a vehicle, and then identifies the detected road signs. This paper presents an automatic neural-network-based road sign recognition system. First, a study of the existing road sign recognition research is presented. In this study, the issues associated with automatic road sign recognition are described, the existing methods developed to tackle the road sign recognition problem are reviewed, and a comparison of the features of these methods is given. Second, the developed road sign recognition system is described. The system is capable of analysing live colour road scene images, detecting multiple road signs within each image, and classifying the type of road signs detected. The system consists of two modules: detection and classification. The detection module segments the input image in the hue-saturation-intensity colour space, and then detects road signs using a Multi-layer Perceptron neural-network. The classification module determines the type of detected road signs using a series of one to one architectural Multi-layer Perceptron neural networks. Two sets of classifiers are trained using the Resillient-Backpropagation and Scaled-Conjugate-Gradient algorithms. The two modules of the system are evaluated individually first. Then the system is tested as a whole. The experimental results demonstrate that the system is capable of achieving an average recognition hit-rate of 95.96% using the scaled-conjugate-gradient trained classifiers.

Kao, Yonggui; Gao, Cunchen

doi: 10.1007/s00521-007-0121-ypmid: N/A

Some sufficient conditions for the global exponential stability of cellular neural networks with variable coefficients and time-varying delays are obtained by a method based on a delayed differential inequality. The method, which does not make use of Lyapunov functionals, is simple and effective for the stability analysis of cellular neural networks with variable coefficients and time-varying delays. Some previous results in the literature are shown to be special cases of our results.

Heh, Jia; Chen, Jen; Chang, Maiga

doi: 10.1007/s00521-007-0115-9pmid: N/A

The neural networks are successfully applied to many applications in different domains. However, due to the results made by the neural networks are difficult to explain the decision process of neural networks is supposed as a black box. The explanation of reasoning is important to some applications such like credit approval application and medical diagnosing software. Therefore, the rule extraction algorithm is becoming more and more important in explaining the extracted rules from the neural networks. In this paper, a decompositional algorithm is analyzed and designed to extract rules from neural networks. The algorithm is simple but efficient; can reduce the extracted rules but improve the efficiency of the algorithm at the same time. Moreover, the algorithm is compared to the other two algorithms, M-of-N and Garcez, by solving the MONK’s problem.

Ghazali, R.; Hussain, A.; Liatsis, P.; Tawfik, H.

doi: 10.1007/s00521-007-0132-8pmid: N/A

Motivated by the slow learning properties of multilayer perceptrons (MLPs) which utilize computationally intensive training algorithms, such as the backpropagation learning algorithm, and can get trapped in local minima, this work deals with ridge polynomial neural networks (RPNN), which maintain fast learning properties and powerful mapping capabilities of single layer high order neural networks. The RPNN is constructed from a number of increasing orders of Pi–Sigma units, which are used to capture the underlying patterns in financial time series signals and to predict future trends in the financial market. In particular, this paper systematically investigates a method of pre-processing the financial signals in order to reduce the influence of their trends. The performance of the networks is benchmarked against the performance of MLPs, functional link neural networks (FLNN), and Pi–Sigma neural networks (PSNN). Simulation results clearly demonstrate that RPNNs generate higher profit returns with fast convergence on various noisy financial signals.