Catálogo de publicaciones - libros

An edge detection method by combining image entropy and Self -Organizing Map (SOM) is proposed in this paper. First, according to information theory image entropy is used to curve up the smooth region and the region of gray level abruptly changed. Then we transform the gray level image to ideal binary pattern of pixels. We define six classes’ edge and six edge prototype vectors. These edge prototype vectors are fed into input layer of the Self-Organizing Map (SOM). Classifying the type of edge through this network, the edge image is obtained. At last, the speckle edges are discarded from the edge image. Experimental results show that it gained better edge image compared with Canny edge detection method.

An approach to network anomaly detection is investigated, based on dynamic self-organizing maps (DSOM) and ant colony optimization (ACO) clustering. The basic idea of the method is to produce the cluster by DSOM and ACO. With the classified data instances, anomaly data clusters can be easily identified by normal cluster ratio. And then the identified cluster can be used in real data detection. In the traditional clustering-based intrusion detection algorithms, clustering using a simple distance-based metric and detection based on the centers of clusters, which generally degrade detection accuracy and efficiency. Our approach based on DSOM and ACO clustering can settle these problems effectively. The experiment results show that our approach can detect unknown intrusions efficiently in the real network connections.

In our previous work we presented a variation of Self-Organizing Map (SOM), CSOM that applies a different learning mechanism useful for situations where the aim is to extract rules from a data set characterized by continuous input features. The main change is that the weights on the network links are replaced by ranges which allows for a direct extraction of the underlying rule. In this paper we extend our work by allowing the CSOM to handle mixed data types and continuous class attributes. These extensions called for an appropriate adjustment in the network pruning method that uses the Symmetrical Tau () criterion for measuring the predictive capability of cluster attributes. Publicly available real world data sets were used for evaluating the proposed method and the results demonstrate the effectiveness of the method as a whole for extracting optimal rules from a trained SOM.

Aiming at the problems happened in the practical application of traditional Principle Component Analysis (PCA), the concept of Relative Principle Component (RPC) and method of Relative Principle Component Analysis (RPCA) are put forward. Meanwhile, some concepts such as Relative Transform (RT), “Rotundity” Scatter and so on are introduced. The new algorithm can overcome some disadvantages of traditional PCA for compressing data when data is “Rotundity” Scatter. A simulation has been used to demonstrate the effectiveness and practicability of the algorithm proposed. The RPCs selected by RPCA are more representative, and the way to choose RPCs is more flexible, so that the application of the new algorithm will be very extensive.

Extracting proper features is crucial to the performance of a pattern recognition system. Since the goal of a pattern recognition system is to recognize a pattern correctly, a natural measure of “goodness” of extracted features is the probability of classification error. However, popular feature extraction techniques like principal component analysis (PCA), Fisher linear discriminant analysis (FLD), and independent component analysis (ICA) extract features that are not directly related to the classification accuracy. In this paper, we present two linear discriminant analysis algorithms (LDA) whose criterion functions are directly based on minimum probability of classification error, or the Bayes error. We term these two linear discriminants as recursive Bayesian linear discriminant I (RBLD-I) and recursive Bayesian linear discriminant II (RBLD-II). Experiments on databases from UCI Machine Learning Repository show that the two novel linear discriminants achieve superior classification performance over recursive FLD (RFLD).

The ability to predict traffic variables such as speed, travel time and flow, based on real time and historic data, collected by various systems in transportation networks, is vital to the intelligent transportation systems (ITS). The present paper proposes a method based on Principal Component Analysis and Support Vector Regression (PCA-SVR) for a short-term simultaneously prediction of network traffic flow which is multidimensional compared with traditional single point. Data from a typical traffic network of Beijing City, China are used for the analysis. Other models such as ANN and ARIMA are also developed as a comparison of the performance of both these techniques is carried out to show the effectiveness of the novel method.

The high non-linearity, serious time-variability and uncertainty result in a number of very challenging problems in working on the monitoring and control of biological processes. Many important variables are difficult to measure during monitoring and control. Software sensors can give estimation to unmeasured state variables according to the measured information provided by online measuring instruments available in the system. This offers an alternative feasible program for online measurement. A hybrid soft measurement model that combines principal component analysis with artificial neural networks is applied to monitor the sequencing batch reactor (SBR) process. Simulation results show that the most unmeasured variables can be predicted and the method can capture the main trend of the data.

Two-dimensional principal component analysis (2DPCA) proposed recently overcome a limitation of principal component analysis (PCA) which is expensive computational cost. Symmetrical principal component analysis (SPCA) is also a better feature extraction technique because it utilizes effectively the symmetrical property of human face. This paper presents a symmetry based two-dimensional principal component analysis (S2DPCA), which combines the advantages of 2DPCA and of the SPCA. The experimental results show that S2DPCA is competitive with or superior to 2DPCA and SPCA.

With independent component analysis (ICA) to realize the blind separation from mixed acoustic objects, a recognition method based on support vector machine/Gaussian mixture models (SVM/GMM) is proposed through extracting linear prediction coefficient (LPC) feature. It is revealed that LPC is consistently better than wavelet energy feature, ICA is efficient algorithm to estimate the unknown signal level. This method uses the output of GMM to adjust the probabilistic output of SVM. The validity of the ICA and SVM/GMM model is verified via examples in mixed acoustic objects recognition system.

In this paper, we propose a new super-resolution face hallucination and recognition method based on Independent Component Analysis (ICA). Firstly, ICA is used to build a linear mixing relationship between high-resolution (HR) face image and independent HR source faces images. The linear mixing coefficients are retained, thus the corresponding low-resolution (LR) face image is represented by linear mixture of down-sampled source faces images. So, when the source faces images are obtained by training a set of HR face images, unconstrained least square is utilized to obtain mixing coefficients to a LR image for hallucination and recognition. Experiments show that the accuracy of face recognition is insensitive to image size and the number of HR source faces images when image size is larger than 8×8, and the resolution and quality of the hallucinated face image are greatly enhanced over the LR ones, which is very helpful for human recognition.